Datasets:

ai-aerospace

/

ams_data_full_2000-2020

like 0

Follow

AI Aerospace 2

395 Conclusions and Lessons Learned The primary lesson learned from this mission goes back to the age old saying of “test early and test often”. Much of the issues in this program stemmed from not doing enough validation testing up front to develop a better understanding of early concepts . A notable lesson from testing early however is that care should be taken to understand the applicable boundary conditions of the system. Early trade studies using both analysis and testing should be carried out with an effort to emulate the final system bou ndary conditions. Early effort test booms in this program experienced several stress ruptures; these were eliminated by including the long- term storage test in the development and qualification test flows. In addition, early hinge torque tests did not incl ude the boom end fittings or even full -length booms. It was found that the parts of the boom outside the hinge regions, especially the proximal end fitting, change the boundary conditions of the hinge significantly and therefore increased the hinge output torque. Without this increased torque the hinges would not have met the torque margin requirements, but the increased torque with the end fitting overcame the resistive torque from the electrical harness. Additional critical lessons learned in this progr am were based on the lack of accepted qualification methods for high strain composite materials. At the start of this program, it was identified that the allowable material properties based on existing standards would not be enough for characterizing the relevant performance properties for thin flexible composite structures. To qualify these structures in a relevant manor the simp le methods described in this paper were devised to satisfy the needs of a high value new space customer. References 1 Adams, D., Mobrem, M., “MARSIS Antenna Flight Deployment Anomaly and Resolution”, AIAA.ASME.ASCE.AHS.ASC Structure, Structural Dynamics, and Materials Conference, Newport, Rhode Isla nd. (2006) 2 Yee, J., Pellegrino, S., “Composite Tube Hinges”, Journal of Aerospace Engineering, October 2005, pp 224-231. 3 Mallikarachchi, H., Pellegrino, “Deployment Dynamics of Composite Booms with Integral Slotted Hinges”, 50th AIAA/ASME/ASDCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, Palm Springs, California (2009) 4 Mallikarachchi, H., Pellegrino, S., “Design and Validation of Thin- Walled Composite Deployable Booms with Tape- Spring Hinges”, 52nd AIAA/ASME/ASDCE/AHS/ASC Struct ures, Structural Dynamics, and Materials Conference, Denver, Colorado (2011) 5 Mallikarachchi, H., Pellegrino, S,. “Optimized Designs of Composite Booms with Integral Tape- Spring Hinges” 51st AIAA/ASME/ASDCE/AHS/ASC Structures, Structural Dynamics, and Ma terials Conference, Orlando, Florida (2010) 6 Pellegrino, S., Kebadze, E., Lefort, T., and Watt, A., “Low -Cost Hinge for Deployable Structures” Caltech, July 2002(http://www.pellegrino.caltech.edu/publications) 7 Fernandez, J., and Murphey, T., “A Simple Test Method for Large Deformation Bending of Thin High Strain Composite Flexures” AIAA SciTech Forum, Kissimmee, FL. (2018). 8 Sanford, G., Biskner, A., and Murphey, T., “Large Strain Behavior of Thin Unidirectional Composite Flexures,” 51st AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, Orlando FL. (2010) 9 Murphey, T., Peterson, M., Grigoriev, M., “Large Strain Four -Point Bending of Thin Unidirectional Composites”, Journal of Sp acecraft and Rockets, 52, Feb 2015, pp 882- 895. 10 Rose, T., K. Medina, W. Francis, K. Kawai, A. Bergan, and J. Fernandez. "Viscoelastic Behaviors of High Strain Composites." In 2019 AIAA Spacecraft Structures Conference. 2019. 11 Sharma, A. H., T. J. Rose, A. Seamone, T. W. Murphey, and F. Lopez Jimenez. "Analysis of the Column Bending Test for Large Curvature Bending of High Strain Composites." In AIAA Scitech 2019 Forum, p. 1746. 2019.

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 397 Mars 2020 Motor Bearing Failure, Investigation and Response David Suffern*, Jeff Mobley* and Stephen Smith* Abstract The prime mover s in each joint of the external robotic arm of NASA Jet Propulsion Laboratory’s (JPL) Mars 2020 rover are planetary gearmotors containing a brushless DC motor and brake assembly , designated as M45S. Qualification l ife testing was performed at multiple levels of assembly in an effort to retire risk to t he program. A significant failure occurred within the motor in the process of executing a life test at the gearmotor level of assembly . The combination of bearing retainer , lubrication scheme, test temperatures , and long life led to a stall failure of the M45S front bearing at 78 million revolutions, not satisfying the life test requirement of 105.2 million motor rev olution s, which is double the expected life. Ultimately, increasing the minimum life test temperature from -70°C to -55°C allowed for succe ssful qualification of the robotic arm joints with the baseline bearing configuration. Details of the requirements, design, life test, test failure investigation, response , and lessons learned will be presented. Application and Requirements Reliable performance of the two- meter -long robotic arm carried by the Mars 2020 rover, shown in Figure 1, is central to the success of this mission. Each joint of this robotic arm is driven by a M45S motor and brake assembly. The M45S was incorporated into three different planetary gear motor designs covering multiple applications: s houlder and elbow (ShEl), w rist and t urret (WAT), and feed. The ShEl and WAT gearmotors were then further integrated into harmonic drive mechanisms in the robotic arm. Each application has unique load and life requirements based on the estimated robotic arm usage and travel. Of these, the ShEl application provides the enve loping requirement for mission life at 52.6 million motor revolutions , resulting in a margined life test requirement of 105.2 million motor revolutions . Figure 1 – Robotic Arm of the Mars 2020 Rover (Courtesy of NASA/JPL- Caltech) * Sierra Nevada Corporation, Durham, NC ; [email protected] Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 398 The M45S has two primary functions: (1) p rovide the required torque and speed when powered and commanded to move, and (2) hold position when unpowered. Motor holding torque capability, necessary for maintaining the robotic arm’s position when the motor is unpowered, was a crucial requirement in the design and development of the M45S motor and brake assembly , but is not covered in this paper . Rather, the operational motor requirements and the failure to demonstrate margined life capability against those requirements will be discussed. In operation, t he M45S motor was required to produce a minimum torque of 0.13 N •m at a minimum speed of 3,00 0 rpm across an operational temperature range of -70°C to +70° C. Given the bias of the Martian environment to colder temperatures and the associated increase in drag torque of the downstream gear ing, a second performance requirement was given for temperatures at or below - 30°C: minimum torque of 0.20 N•m at a minimum speed of 2,400 rpm. These two performance points were the basis of the motor design and sizing. Additional design goals and test requirements regarding operation at colder temperatures increased the difficulty of the program. Specifically, there was a program goal to minimize torque variation over the wide operational temperature range. A limited tor que uncertainty across operating conditions would allow JPL to calculate output torque produced by the motor solely as a function of the current input. In real terms, this translated to design direction to avoid adding excessive lubricant to the motor bear ings, understanding that additional lubricant would increase the motor drag and therefore the torque uncertainty at colder temperatures. Additionally , one of the most challenging test requirements, necessary to represent the bias of operation at colder t emperatures, was to satisfy the following distribution of temperatures in the execution of the margined life test (105.2M motor shaft revolutions ): • Minimum 25% of the revolutions at the hot extreme (+70° C). • Minimum 25% of the revolutions at the cold extreme ( -70°C). • Minimum 25% of the revolutions at the nominal operating temperature ( -55°C). Design and Evaluation The M45S motor and brake assembly wa s comprised of a brushless DC motor with Hall commutation and a fricti on brake assembly . The rotor at the core of the M45S was supported by two radial ball bearings, spring preloaded in a face- to-face orientation . Each double- shielded, SR3 size bearing contained a crown steel retainer and was lubricated with a grease plate of Braycote 600EF followed by the addition of a 1:1 by volume slurry mixture of Braycote 600EF grease and Brayco 815Z oil to fill 5% to 10% of the bearing’s void volume. The quantity of lubricant was minimized in order to keep torque losses low, and to reduce drag torque variation over temperature extremes. It should be noted that the combination of a grease plate prior to the addition of 5% to 10% fill of 1:1 grease/oil slurry resulted in a total lubricant condition of 8% to 13% volume fill and a grease/oil ratio closer to 2:1. This lubrication approach was selected based on its heritage success [1] [2] [3] . Additionally, the crown steel retainer had heritage, being successfully used in Mars Science Laboratory gearboxes and in developmental testi ng for Mars 2020 gearbox designs . However, the combination of this retainer configuration, the lubricant fill, and the high number of revolutions required to be performed at the margined operating temperature, - 70°C, pushed this bearing configuration beyond the acceptable limits of operation. This will be explained in detail. A simplified cross -section of the M45S motor is shown in Figure 2.

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 399 Figure 2 – Simplified M45S Motor Cross -Section (Brake omitted for clarity) The rotor magnets overhang the rear of the stator to allow efficient Hall sensor commutation. This overhang increas ed the load on the front motor bearing due to the magnetic forces attempt ing to self -center the rotor within the stator. As a result, Hertzian contact stresses were expected to be higher [0.99 GPa (144 ksi) ] in the front bearing and lower [ 0.80 GPa (116 ksi) ] in the rear bearing. These are reasonable values for a long life application. The bearing lubrication film parameter ( λ), a ratio of the lubrication film thickness to the composite roughness of the contacting surfaces between the ball and raceway , was calculated using COBRA -AHS analysis software across temperatures ranging from -70°C to +70° C and speeds ranging from 1, 500 rpm to 7,100 rpm. The analysis indicated a mixed boundary condition λ of 1.7 at slower speeds and higher temperatures , and a hydrodynamic (HD) condition (λ >3) for higher speeds and colder temperatures [4] [5]. The lubricant ’s viscosity is a function of temperature and increases at colder temperatures. This increase in viscosity inherently increases the lubrication film thickness, which i n general, results in longer bearing life as the lubricant is able to insulate the bearing elements from direct contact [6] . However, a simple extrapolation of lubricant viscosity going cold may be insufficient for understanding impact on bearing life. As will be shown in Figure 10, the lubricant ceases to function effectively when approaching its pour point , resulting in starvation effects that dramatically reduce bearing life. In addition to Hertzian contact stress and the lubrication film parameter, lubricant stress cycles are us ed in evaluating bearing life. These are defined as the number of times a ball passes across a gi ven spot on the raceway, a function of the bearing’s pitch diameter, ball diameter, and quantity of balls. See Eq. 1 in reference [2] for calculation details and an example in reference [5]. The lubricant stress cycles for the bearings in the M45S motor were 469 million for the margined life requirement of 105.2 million motor revolutions . Published and other available data summari zing acceptable limits of Hertzian contact stress and lubricant stress cycles indicated that the M45S design was exceeding limits of proven lubricant life, but not in an area of known failure [3] [7]. The data indicate s that at least 30-60 million lubricant stress cycles could be expected for contact stresses around 1.0 GPa (145 ksi), with failure to be expected beyond 600 million lubricant stress cycles . Being between the limits of proven success and known failure, bearing lubricant life was identified as a significant program risk throughout the design phase. In response, the bearing balls were coated with titanium carbide (TiC), shown to quadruple the life of the perfluoropolyether (PFPE) lubricant at Hertzian contact stresses of 1.0 GPa or less [8]. Concurrent with the M45S design effort , developmental bearing life testing was being performed on similar bearings being used Stator Rotor Hall Effect Sensors Motor Housing Preload Spring Rear Bearing Front Bearing

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 400 on a separate Mars 2020 cont ract to sup ply gearboxes. This testing was intended to demonstrate bearing life capability in the low cycle count , high load regime [ 2]. Unfortunately, i t was not possible to adapt that test setup to encompass the speeds or cycle count required for the M45S motor design. Consequently, demonstration of M45S bearing life was deferred until its design verification (DV) life testing . Performance and Life Test Results M45S Performance The goal to minimize torque variation over temperature was a significant program concern and drove the decision not to implement a higher bearing lubricant fill . Performance testing was completed and proved that the chosen motor design, including bearing lubrication, performed consis tently across the entire operational temperature range. The net motor torque constant , including viscous losses from lubrication, across a temperature range of -70°C to +70 °C resulted in a torque uncertainty of ±0.005 N•m at the 0.200 N•m output torque lev el at constant speed. This value was within the original program goal of ±0.009 N•m . With respect to torque uncertainty concerns, t he chosen design and lubrication scheme proved highly effective. Life Testing and Results Identified early in the program as an area of risk, the lubricant life w as ultimately tested during the M45S DV life test, which exposed the unit to twice the expected life, or 105.2 million motor revolutions. This life test, completed in early 2019, had been delayed due to other M45S design and integration challenges . These delays postponed the retirement of the bearing life risk. As a result, within six weeks of finishing the M45S DV life test, two of the M45S gearmotor life tests ( ShEl DV and WAT DV ) were completed or stopped due to anomalous conditions. This timing afforded the opportunity for near -concurrent visual inspection of motor bearings experiencing the same operational temperature environment, but different total motor revolutions with various speed and load combinati ons. The revolution count and condition of the M45S motor bearings from these three initial life tests were compared for similarities and differences, shown in Table 1. Table 1 – Results of Life Tests Ending in Early 2019 Life Test Motor Revs / % of 2X Life Temperature Distribution Post -Life Performance Visual Bearing Condition M45S Motor -only 105M / 100% 25% min. @ - 70°C 25% min. @ - 55°C 25% min. @ +70 °C Acceptable Poor WAT Gearmotor 66M / 100% Acceptable Marginal ShEl Gearmotor 78M / 74% N/A – Stalled Very Poor / Damaged Figure 3 provides a visual comparison of the condition of the front bearing from the M45S mot or of each of these three units, with the following observations: • Motor -Only: Slight ly rough rotation; dry, powdery, dark wear debris; retainer ball pockets worn • WAT : Rough rotation; clumped, dark wear debris • ShEl : Very rough rotation; retainer periodically caught under balls; nearly empty of any lubricant or wear debris; d amage to retainer vis ible

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 401 Figure 3 – Post-Life Test Condition of Front M45S Motor Bearings Details of the ShEl Life Test and Termination The program required that life tests perform twice the expected number of Integration and Test (I&T) and Mission revolutions over at least 10 thermal cycles, each consisting of temperature plateaus at +70° C, -70°C, -55°C, and +20° C. Additionally, the gearmotor output revolutions were required to be approximately equally distributed across the thermal cycles with a minimum of 25% of the revolutions to be performed at each of the +70° C, -70°C, -55°C plateaus . The balance of revolutions could be performed at any temperature between +70° C and - 70°C, but were mostly completed around +20 °C. By early 2019, nearly all of the gearmotor DV life tests had been completed without issue. However, the ShEl life test unit unexpectedly stopped operating during the - 55°C plateau (just after the - 70°C plateau) of thermal cycle 6 of 10 in early 2019. Figure 4 provides an overview of the data acquired during this plateau, including the three stall events that occurred prior to halting operation for evaluation. The stall events are indicated by the t hree sharp rises in the motor current and temperature followed by recovery . Figure 4 – Output Torque and Motor Current Data Acquired During Thermal Cycle #6, - 55°C Plateau

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 402 Figure 5 provides detailed data of the first stall event as i t occurred, with the following explanation: • Per the automated profile, the unit bega n execution of 80 seconds of operation in the CW direction. • After approximately 15 seconds of operation against the prescribed torque of approximately 6.6 N•m, the measured output speed of the unit unexpectedly dropped to 0 rpm and the motor current jumped to the 7 A current limit. • The unit remained in this state for approximately 60 seconds , until the automated profile removed power from the motor . The motor remain ed off for 15 to 20 seconds, as planned. • The automated profile initiate d operation in the CCW direction. • Operation then proceed ed nominally for the next 80 to 90 minutes until a second stall event was encountered with the same signature as the first (reference Figure 4). Note, some initial torque noise is visible on the graph and is associated with t he cogging effects of the hastily stopped hysteresis brake dynamometer . All three stall events were readily identified by the sharp increase in motor housing temperature associated with Joule heating from the motor current being at its 7 A limit. Figure 5 – Detailed Data Acquired during First Stall Event of ShEl DV in Life Cycle 6 of 10 Test Failure Investigation The team first needed to determine whether the stall was due to an anomaly within the unit under test (UUT) or the mechanical or electrical ground support e quipment (MGSE / EGSE) . The life test setup was more complicated than most due to the requirements to apply resistive torque and axial load simultaneously, while support ing the UUT in the thermal chamber and recording data as noted. The complexity of the test setup presented multiple opportunities for the root cause to be found in the MGSE / EGSE. A block diagram of the complicated ShEl life test setup is shown in Figure 6, with data acquired from components identified in the orange boxes .

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 403 Figure 6 – Block Diagram of Complex ShEl DV Life Test Setup The following describes the sequence of the investigation to determine the source of the ShEl DV unit life test anomaly : 1. To determine if the EGSE was responsible, t he team attempted to recreat e the stall event while recording the current of each motor phase and the voltage of each Hall sensor with an oscilloscope. Result: stall was recreated and oscilloscope data indicated EGSE was performing nominally. 2. The performance of the motor’s brake was characterized to ensure it was not inadvertently engaging and stopping the motor. Result: motor brake performance was verified to be nominal. 3. Thermal chamber was taken to ambient temperature and a recreation of the stall event w as again attempted in a manner that would allow a determination of the source of the anomaly: MGSE or UUT. Result: stall was recreated and there was no wind- up or tension in the MGSE test setup to indicate that it was causing any issues. 4. UUT was disconnect ed from all MGSE and a Startup Sensitivity health check was attempted to verify the motor was able to initiate rotation at a current value in family with prior health checks . The UUT failed to rotate with up to 2.8 A applied, when far less current was gene rally needed to initiate rotation. This provided convincing evidence that the source of the stall was within the UUT itself. 5. JPL performed a computed tomography (CT) scan on the motor and discovered that the source of the stall condition was within th e front motor bearing; the crown steel bearing retainer jammed under a ball , as shown in Figure 7. Figure 7 – CT Scan of Front M45S Bearing within ShEl Life Test Unit (Courtesy of NASA/JPL- Caltech)

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 404 The UUT was returned and disassembled at SNC. T he interior condit ion of the motor , particularly the front bearing, was evaluated by both SNC and JPL personnel. The following observations were made regarding the condition of the ShEl gearbox and motor: • No significant anomalies were found within the ShEl gearbox . • No significant anomalies were found within the M45S brake assembly . • In disassembly of the motor, the r ear bearing rotated smoothly and residual, dark , and slightly wet lubricant remained. • A significant amount of wear debris was present on both sides of the fr ont motor bearing: motor rotor and motor pinion. See Figure 8. • The f ront bearing rotated rough ly and caught occasionally. Two fingers of the ball retainer were visibly damaged. The bearing was v ery dry and empty of both lubricant and wear material (appeared to have been ejected into motor rotor and motor pinion areas). The inner (piloting) diameter of the crown steel retainer inc luded a rolled burr. Further inspection found that this inner diameter was measurably enlarged, allowing radial movement of the retainer to the point where the outer diameter of the retainer could contact the inner diameters of the outer race. Additionally , there were indications that the retainer spine contacted the interior surface of the bearing shield , indicating that the retainer ball pockets had worn to allow excessive axial movement of the retainer . See Figure 9 for the appearance of this damaged front bearing. Figure 8 – Wear Debris Surrounding the M45S Front Motor Bearing from ShEl DV Unit Figure 9 – Interior Inspection and Retainer Observations on M45S Front Motor Bearing from ShEl DV Unit

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 405 The results of this and other unit teardowns led to the development of the following failure sequence theory : 1. Operation at the - 70°C temperature plateau, very close to the lubricant pour point of - 72°C, rendered the lubrication scheme ineffective, resulting in lubricant starvation conditions. 2. Starvation conditions , never a desir ed condition for long life bearing operation, had the most significant effects on the crown steel retainer, wearing first at the inner diameter (ID) of bearing retainer spine, where guided. 3. Retainer w ear debris mixe d with grease , accelerating wear as the grease became an abrasive mixture with the wear particles. 4. Retainer ID continued to grow with wear, allowing it to move radially out of position. a. Ball pockets began to wear against balls, generat ing debris and burrs . b. Outer diameter faces of the retainer began to wear on the outer ring and inside of shields . 5. Wear debris led to increased drag in the bearings , causing increased noise in the motor current signature. 6. Eventually lubricant was depleted and dry debris was ejected from the bearing . 7. Two types of failure were observed: a. The bearing retainer was jammed between the bal l and outer ring, stalling the motor. This mode was the initial life test failure observed in the original ShEl DV M45S motor. b. The b uild-up/plating of wear particles and depleted grease on raceways removed radial play from the bearing and caused a significant increase in drag/motor current, but not an outright motor stall. This failure occurred in the M45S SN013 motor that had replaced the original ShEl DV motor to complete the margined gearbox life test. Response to M45S Bearing Failure This life test failure occurred on one of the last gearmotor DV life tests and after the delivery and integration of the flight gearmotors into the Mars 2020 rover’s flight robotic arm. The state of the program and the severity of the issue resulted in parallel recovery paths being pursued: 1. “Use as is” – Determine an operating tempe rature range that would allow the baseline, delivered bearing configuration to achieve the margined life test requirement . The goal was ensure the lubricant was able to flow and protect the crown steel retainer from lubricant starvation wear. 2. “Refurbish” – Determine a superior bearing configuration that w ould achieve the margined life test requirement without failure. Adjust operating temperature range as necessary to maintain acceptable torque uncertainty. “Use as is” was t he option with least impact to t he assembly and integration work already accomplish ed by the Mars 2020 rover team. However, the refurbishment option was simultaneously pursued as a backup plan in the event there was no reasonable operating temperature range that allowed the delivered hardware configuration to achieve the margined life test requirement . SNC and JPL had a number of M45S motors available as test beds for testing various bearing configurations across various temperature ranges . In addition, SNC possessed a number of motor controller s, test consoles , and thermal chambers to support simultaneous testing. Among many options, the collective program team considered the following variables valuable t o compare: • Minimum mission life test temperature • Speed • Lubricant mix ratio and percentage of fill • Retainer material Within three weeks of determining the front mot or bearing to be the cause of the ShE l DV life test stall, the JPL-SNC team had begun a modified life test on two available M45S motors of the baseline bearing configuration. B oth the flight spare motor (SN019) and a non- flight thermal correlation motor (SN032) began life testing , alternating between two less severe temperat ure plateaus: - 35°C and +25° C. To increase the fidelity of these life tests and match the test flow , the motors were expose d to the random vibration

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 406 environment prior to the beginning of life testing. Although they had the s ame baseline bearing configuration, SN032 performed its life test at the high end of the operational speed (7000 rpm), and was completed within three weeks , whereas SN019 performed its life test at the low end of the operational speed (1500 rpm), and took two months to complete. In reality, the actual motor speed in mission operation on the rover could be at any speed in between. These two speeds were selected to bound mission operation, since speed could have a potentially significant impact in the bearing lubrication film parameter and overall bearing life. In parallel to these efforts, the program team chose to procure an alternate crown phenolic retainer in the “refurbish” path in lieu of the baseline crown steel retainer. This decision was made due the determination that the root cause of failure was the compounding deterioration of the crown steel retainer in the absence of effec tive lubrication at cold temperatures. The premise was that the porous phenolic retainer had more published heritage success and itself was a potential lubricant reservoir, inherently resistant to lubricant starvation. To this end, JPL was able to successf ully modify four bearing sets to include crown phenolic retainers in place of the crown steel version, each set having a different lubrication scheme, as follows: • 58.1 mg ( 30% void volume fill) of Braycote 600EF grease only • 38.7 mg ( 20% void volume fill) of Braycote 600EF grease only • 30 mg ( 15.5% void volume fill) of 2:1 slurry of Braycote 600EF grease : Brayco 815Z oil • 22 mg ( 11.5% void volume fill) of 2:1 slurry of Braycote 600EF grease : Brayco 815Z oil The purpose of the various lubrication schemes was to compare cold temperature performance. This performance test ing was used to determine the drag of each bearing configuration with respect to temperature in order to assess the torque uncertainty associat ed with these alternate configurations. The JPL Motion Control Subsystem (MCS) team evaluated the results of this testing with respect to torque uncertainty goals . The cold performance test results for these lubricant configurations are shown in Figure 10, with the following observations: • SN012 motor with 58.1 mg grease reached the 0.6 A current limit and could not operate at or below -50°C. Even at warmer temperatures , the no- load current was above 0.5 A and this lubricant configuration was deemed unusable by the JPL MCS team. • The other three configurations had acceptable performance but only the SN016 motor with 38.7 mg grease and SN012 with 22 mg slurry proceeded to life testing. • The data show ed a low torque characteristic at temperatures at or below -60°C. This characteristic is believed to be from a starvation condition where the lubricant is so close to its pour point that its high viscosity does not allow for it to effectively flow within and relubricate the bearing components . While this produces low drag torque, it is not acceptable for a long life application. • Peak torque was found to occur around -50°C indicating that the lubricant viscosity was still high due to the low temperatures but was able to flow within the bearing.

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 407 Figure 10 – M45S Cold Performance with Phenolic Bearing Retainer and Various Lubricant Schemes The timeline of recovery for the M45S motor, including assembly, test, and teardown inspection events is shown in Figure 11 with the following details: • Testing, including random vibration, began with the baseline bearing configuration in SN019 and SN032, modifying the operating temperature range to -35°C to +25° C to be less severe than the original -70°C to +70° C. • Refurbishment of SN012 and SN016 was completed after receiving bearings with phenolic retainers and the noted grease fill from JPL. o Cold temperature drag testing revealed that the bearings in SN012 with 58.1 mg grease fill had excessively high drag. That unit, therefore, was not selected to continue to life testing. o SN016 did proceed to life testing over the - 35°C to +25 °C temperature range in the event that the SN032 with the baseline bearing configuration f ailed. • SN032 life testing successfully completed with favorable results over the -35°C to +25° C temperature range, indicating that the baseline bearing configuration was capable of meeting the margined life requirements with a minimum operating temperature of -35°C. o As a result, life testing of SN016 was halted and the unit was made ready for a different set of bearings. By this point, SN016 had accumulated 75 million revolutions (71% of margined l ife). • With successful results of the baseline bearing configuration at -35°C, the program set out to determine if either the baseline or a refurbished bearing configuration could successfully meet the margined life requirements with a lower minimum operating temperature of - 55°C, still warmer than the original requirement of - 70°C. A lower minimum operating temperature would reduce the needed motor heater power, conserving the rover’s limited power supply for mission operations . This testing was performed with three units : o The slow speed , long -duration SN019 life test , already in progress , was modified to make up ground at the -55°C plateau. o SN016, refurbished to include the baseline bearing configuration, was run at full speed. o SN012, refurbished with bearings JPL modified to include phenolic retainers and the 22-mg slurry fill, was run at full speed as well . • SN012 and SN016 were exposed to both the random vibration environment and the Planetary Protection (PP) Bake- out, which require d at least 122 hours at + 114°C and a pressure of less than 1x10-5 torr. The bake- out was included to ensure that the bearing lubricant would be conditioned in a manner consistent with the flight motors. • All three motors finished testing within a week of each other. A Technical Interchange Meeting (TIM) was held to review all motor performance data and visual bearing condition s.

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 408 Figure 11 – Timeline of M45S Motor Recovery Assembly, Test, and Teardown Events In the TIM, the decision was made to proceed with the baseline bearing configuration for the qualification and flight mechanisms, provided the minimum mission life test temperature was increased from -70°C to -55°C. Furthermore, since the M45S was capable of meeting performance requirements at colder temperatures and the concern was regarding reduced mechanism life, it was recommended that the power allocat ion for robotic arm motor heating be in proportion to the exp ected life of each joint . This allocation would conserve limited rover power resources while maximizing overall robotic arm life. Practically, this means that a joint that is expected to experience minimal revolutions could be operated colder than a joint that is expected to experience the most revolutions. Additionally , it was recommended that heater power continue to be used as available to maximize life , not just to achieve a minimum - 55°C operating temperature. Motor performance was acceptable for all five of the M45S recovery test configurations. A summary of the variables tested and the outcome of those tests are shown in the matrix of configurations and environments in Table 2: Table 2 – M45S Motor Bearing Recovery Test and Results Matrix Motor SN Retainer / Lubricant I&T Temp erature / Motor Revolutions Mission Temp erature / Motor Revolution s Total Revs Speed (rpm) Visual Cond . SN032 Crown Steel 1 -33°C +25°C 106M 7000 OK 53.0M 52.8M SN016 Crown Phenolic 2 -33°C +25°C 76M 7000 Good 42.4M 33.1M SN012 Crown Phenolic 3 -70°C +25°C +70°C -55°C -35°C +25°C 106M 7000 Good 1.86M 3.88M 1.79M 39.3M 29.5M 29.9M SN016 Crown Steel 1 -70°C +25°C +70°C -55°C -35°C +25°C 106M 7000 OK 1.86M 3.87M 1.79M 39.2M 29.4M 29.8M SN019 Crown Steel 1 -70°C +25°C +70°C -55°C -35°C +25°C 105M 1500 OK 1.85M 3.85M 1.81M 39.2M 29.4M 29.3M 1 Baseline Lubrication: Grease Plate of Braycote 600EF Grease with 5% to 10% fill with 1:1 slurry by volume of Braycote 600EF grease and Brayco 815Z oil 2 No grease plate. Filled with 38.7 mg Braycote 600EF Grease, or ~ 20% fill 3 No grease plate. Filled with 22 mg of 2:1 slurry of Braycote 600EF grease: Brayco 815Z oil , or ~11.5% fill

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 409 Figure 12 is a comparison of the post-life test condition of the front motor bearing from various units: • Left: O riginal M45S DV motor -only life test with mission temperatures of - 70°C to +70° C. Bearing spun roughly and the lubricant/wear debris was dry and dark. • Center: M45S SN016 full-speed baseline bearing life test with modified mission temperatures of -55°C to +25° C. Bearing spun smoothly and lubricant was still present and slightly discolored around the retainer ball pocket areas. • Right: M45S SN012 full-speed life test with modified mission temperatures of - 55°C to +25° C and a phenolic retainer with 22 mg grease/oil slurry fill . Bearing spun smoothly and lubricant was plentiful and minimally discolored. Figure 12 – Post-Life Condition of Front M45S Bearing s Conclusion s and Lessons Learned Conclusions While the end- of-life condition of the bearings with phenolic retainers was superior to thos e with baseline crown steel retainers, the level of risk and schedule impact associated with replacing bearings installed in flight motors was deemed too significant. Multiple life tests confirmed that increasing the minimum mission life test temperature from - 70°C to -55°C result ed in the baseline bearing configuration meeting the margined life requirements. The team planned to implement this change in minimum operating temperature by allocating the motor heating power budget proportionally based on the life requirement of each M45S motor application. By October 2 019, the robotic arm joint assemblies , using SNC gearmotors containing the M45S motor with the baseline bearing configuration , were qualified for flight after successfully completing their margined life tests utilizing an increased minimum mission life test temperature of -55°C, providing additional confidence in the chosen solution . Lessons Learned • Programs should evaluate the appropriateness of applying a 15°C thermal uncertainty margin in combination with the 2X margin on required revolutions during life testing, especially when the temperature margin results in a significant viscosity change in the lubric ant. • Given the - 72°C pour point of Brayco 815Z oil , operation at - 70°C result ed in a starvation condition. Future work should evaluate newer lubricant formulations with lower pour points if long duration operation at - 70°C is unavoidable. • Bearing drag t esting over temperature may be useful as a means of determining the low -end temperature limit for effective wet lubrication. Bearing drag normally increases with decreasing temperature, but drag torque was shown to decrease below this low -end temperature l imit as the lubricant cannot flow and the bearing operates in a lubricant starvation condition.

Document

AMS_2020.pdf

© 2020 Sierra Nevada Corporation 410 • Bearing lubricant life depends on many factors in addition to industry accepted Hertzian contact stress and lubricant stress cycles. A developmental life test of the flight configuration of bearings is recommended as early as possible to ensure succes s while accounting for as many other variables as possible: bearing retainer design and material, lubricant fill, test temperature and duty cycle, film thickness at design speed(s) , etc. • Engineering should identify and resolve conflicting design requirements early in the program: low drag torque and long life requirements are potentially in conflict with a wet lubrication scheme. • In this design and environment, phenolic retainers would have provided longer life than crown steel retainers. Acknowledgements The authors wish to express appreciation to the Motion Control Subsystem (MCS) team at JPL for their collaboration in this investigation and resolution. Specific contributors of importance were Mark Balzer, Andrew Kennett, and Alex Ferreira. This work was performed for the Jet Propulsion Laboratory, California Institute of Technology, under the Prime Contract NNN12AA01C between Caltech and NASA under subcontract number Subcontract 1541135. Government sponsorship acknowledged. References 1. Herman, J. and Davis, K., “Evaluation of Perfluoropolyether Lubricant Lifetime in the High Stress and High Stress -Cycle Regime for Mars Applications”, Proc. 39th Aerospace Mech. Symp., NASA CP2008- 215252, Huntsville, AL (2008) 2. Suffern, D. and Parker, J., “Developmental Bearing and Bushing Testing for Mars Gearboxes”, Proc. 44th Aerospace Mech. Symp. , NASA/CP- 2018- 219887, Cleveland, OH (2018) 3. Conley, P. L. and Bohner, J. J., “Experience with Synthetic Fluorinated Fluid Lubricants”, Proc. 24th Aerospace Mech. Symp., NASA CP- 3062, Kennedy Space Center, FL (1990) 4. Zaretsky, E . V. “Lubricant Effects on Bearing Life,” NASA -TM-88875, December 1986 5. Mobley, J.; R obertson, M.; Hodges, C., “Extended Life Testing of Duplex Ball Bearings”, Proc. 43rd Aerospace Mech. Symp. , NASA/CP- 2016- 219090, Santa Clara, CA (2016) 6. Zaretsky, E . V., “STLE Life Factors for Rolling Bearings.” STLE SP -34, Park Ridge, IL, 1999. 7. Lo, C. J., et al. , “Use of Cumulative Degradation Factor Prediction and Life Test Result of the Thruster Gimbal Assembly Actuator for the Dawn Flight Project.” NASA/CR -2009- 215681. 8. Jones, W. R., et al., “The Effect of Stress and TiC Coated Balls on Lubricant Lifetimes Using a Vacuum Ball -on-Plate Rolling Contact”, Proc. 33rd Aerospace Mech. Symp., NASA/TM -1999 -209055, Pasadena, CA (1999)

Document

AMS_2020.pdf

411 Mars 2020 maxon Commercial Motor Development from Commercial -Off-the-Shelf to Flight -Qualified Motors, Gearbox es, and Detent Brake s: Overcoming Issues and Lessons Learned Michael LoSchiavo*, Robin Phillips**, Rebecca Mikhaylov * and Lynn Braunschweig** Abstract Building on previous collaborations, maxon and the Jet Propulsion Laboratory ( JPL) established a partnership to modify Ø20- mm & Ø32- mm Commerc ial-Off-The-Shelf (COTS) BrushL ess Direct Current (BLDC) flat motors and a Ø22- mm planetary gearbox. The commercial design was modified to meet the requirements for the Mars 2020 rover, a Class B [1] interplanetary mission operating in the Martian surface environment , while maintain ing as much of the industrial heritage as possible . Numerous i ssues encountered during development were successfully addressed, and qualification of the design and acceptance testing for the 10 F light Model (FM) actuators installed on the Mars 2020 rover was completed on time for the rover assembly schedule. Mars 2020 Overview and Previous Missions maxon brushed motors have significant flight heritage on JPL Mars missions . The Mars Pathfinder Sojourner rover contained 11 brushed direct current ( DC) motors based on the RE16 d esign [2]. The Mars Exploration Rover s (MER ), Spirit and Opportunity , each utilized 39 maxon brushed DC motors, based on modified commercial “RE” series designs. At the end of the Opportunity rover’s mission , all of the drive wheel motors were still functioning, having survived 5111 Sols (Martian day/night thermal cycles) , with each motor complet ing ~9 × 107 revolutions . Spare MER motors (4× RE20 and 5× RE25) were also used as part of the Phoenix lander. m axon has continued to develop brushed motors for the Martian applications Interior Exploration using Seismic Investigations, Geodesy and Heat Transport (Insight ) mission and ExoMars mission by modifying newer commercial designs , however , these designs were not viable options for Mars 2020 due to the brush sensitivity to a pyroshock environment. Mars 2020 is the first step in a potential multimission approach to a Mars Sample Return mission and future human exploration of Mars [ 3,4]. The primary objectiv e of the Mars Science Laboratory (MSL) and its Curiosity rover is to search for habitablity . Mars 2020 [5] and its Perseverance rover will seek signs of past life and collect Martian cores for possible later collection and return to Earth. The Mars 2020 project maximize d the use of build- to-print hardware to reduce development risk and much of the project utilizes a large amount of spare MSL hardware (specifically the cruise and descent stages) . Mars 2020 also levied design constraints to remain within the MSL footprint in terms of mass, volume , and power on major components , such as the rover chassis, which houses the Sample Caching System (SCS) , and large robotic arm. The SCS is one of the most advanced robotic systems ever developed for planetary exploration, pushing more robotic Degrees of F reedom (DOFs) into a smaller volume than has previously been achieved. Within the SCS, the Adaptive Caching Assembly (ACA) and Coring Drill [6,7,8,9,10] requires complex mechanisms that need gearmotors smaller in volume than any equivalent hardware developed for MSL. Additionally, solenoid brake channels in the motor controller were traded for additional sensor channels, which required any new motors to utilize a passive holding torque mechanism. Tasked with Mars 2020’s mission objective to drill, capture, and store core samples along with volume constraints from the MSL * Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA ** maxon international ag, Sachseln, Switzerland Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center , 2020

Document

AMS_2020.pdf

412 rover , the project re quired new actuators that then needed to be baselined from proven/existing technologies to meet the launch schedule. The Curiosity r over had major actuator development issues , which contribut ed to the launch slipping from 2009 to 2011. The development of several new motors was highlighted as a major reason for the actuator schedule slips [4,11]. The JPL Mars 2020 actuator team considered the lessons learned from the earlier rover mission and decided to leverage an existing COTS design to reduce the flight hardware schedule risk. maxon has an extensive catalog of motors for industrial use and has proven flight heritage from the MER mission that, with suitable modifications, the motor designs are robust for space applications. maxon COTS flat motors were utilized in the SCS testbeds and baselined in the original derivation of the system volume and functional requirements. A close collaboration between maxon and JPL was essential in developing an up-screened COTS -to-robust , flight flat motor mounted on a $2B mission. Two motor types, referred to here as M20 and M32 ( with 2 0 and 32 being the COTS motor diameter in millimeters) , were based on existing maxon COTS products and chosen to meet the project envelope constraints (Figure 1). The M20 gearmotor include s an integrated gearbox (Figure 1) . Additionally, a detent module was added to the M32 to provide a passive holding torque to a static motor. When the motor is spinning above a minimum speed, the inertia is sufficient to overcome any subtractive/additive torques from each detent step, with no significant performance reduction from the motor. Figure 1. Top Left: M32 motor cut -away showing detent brake; Top Right : M20 FM gearmotor ; Bottom: Exploded view of M20 gearmotor Mars 2020 Mechanism Descriptions The SCS includes a rotary percussive drill located on the turret at the end of the large robotic arm and the sample tube processing plant internal to the rover (the ACA, Figure 2). A 3-DOF arm in the processing plant passes the tubes and samples to multiple science stations and hermetically seals the tubes. Additionally , a bit carousel passes tubes and bits between the internal ACA and the external drill. Several of the mechanisms in the SCS subsystem are driven using the M32 detent motors described in this paper, which are mated to three different torque amplifying planetary gearboxes to provide a family of actuators ( named

Document

AMS_2020.pdf

413 SHACD, SAS , and Chuck) , designed and manufactured by Sierra Nevada Corporation in Durham, NC. The motors provide the active driving torque and holding torque for the following applications: the 3- DOF Sample Handling Assembly (SHA) , which consists of a linear stage and two rotary (shoulder/elbow) joints; the tube gripper and sealing ram within the tube sealing station of the ACA; the central rotating axis of the bit carousel; the mechanism that shifts the coring drill between high- speed drilling and high- torque core break - off modes; and the opening and closing of the chuck at the end of the drill. All of these mechanisms are single- string devices performing a serial task in the se quence of capturing a rock core sample. The M20 gearmotor utilized similar modifications as developed in the M32 design effort into the smaller motor diameter for the most volume constrained component of the ACA, the end effector. This component is mounte d to the end of the SHA and physically grabs the sample tube for transport within the ACA. Figure 2. The Adaptive Caching Assembly (ACA) The end effector has some of the strictest planetary protection and handling requirements ever levied on a mechanism. The Mars Helicopter technology demonstrat or mounts to the underbelly of the rover in its stowed configuration and has a one- time actuation of the deploy ment arm. A summary of the four new designs utilizing the M20 and M32 motors for 10 flight locations o n the rover are listed in Table 1. Table 1. Gearmotor assignments Application Gearmotor Name Motor SHA Elbow SHACD M32 SHA Linear SHACD M32 SHA Shoulder SHACD M32 Bit Carousel SHACD M32 Tube Drop -off SHACD M32 Sealing Station SAS M32 Core -Break Lock Out SAS M32 Helicopter Deploy ment Arm SAS M32 Chuck Chuck M32 End Effector DEE (M20) M20

Document

AMS_2020.pdf

414 Mars 2020 Specific Requirements There are several key and d riving requirements derived from the Mars environment and mission requirements : 1. The mission objective to look for compounds that provide evidence of life on Mars levied strict Contamination Control and Planetary Protection handling , design, and assembly constraints , including: reducing organic materials used for assembl y and treating them to minimize outgassing and deposition on crucial components, cleaning all surfaces to prevent contamination from migrating due to handling, and eliminating/reducing viable compounds from spores and organic materials from Earth. 2. The actuators located on the extremities of the rover arm are subjected to wide diurnal temperature changes , which is highly stressing on electronic components and bonded materials. 3. In addition to pyrotechnic release devices, the rotary percussive drill generates an environment that can be defined as a combination of a random vibration, sine vibration, and pyroshock. This environment varies depend ing on several factors , including the hardness of the rock being drilled into, the percussion frequency used for drilling, and the weight -on-bit applied. Ref. [7] contains more information on the coring drill . The dynamic environment fatigues components and can amplify natural modes in hardware, which then couples with fatigue induced from extreme ther mal cycling. maxon Catalog Flat Motor and Gearbox Design maxon has 50 years of experience designing and manufacturing brushed and brushless DC electric motors, gearboxes (both planetary and spur), feedback devices , and control electronics. The brushless DC motors are offered in several families of types, covering both ironless winding and iron- core winding types. The iron-core winding types are in turn divided into external and internal rotor families where the rotor mounted permanent magnets are respecti vely outside or inside the windings. Both the M20 and M32 motors are derived from maxon’s “flat motor” family (Figure 3) , which are of the BLDC external rotor type. For industrial applications, compared to other BLDC types, the flat motors represent a simple, low -cost design that delivers high torque in a short (but large diameter) envelope. When operated without a housing in Earth atmosphere conditions, the exposed spinning outer rotor also enables excellent convective cooling and hence high continuous power ratings. maxon’s flat motor family covers various diameters (14, 20, 32, 45, 60 , and 90 mm), all of which share the same basic design. A large -diameter rotor provides the structural support and acts as a magnet return for a ring magnet. This rotor is supported by two ball bearings housed in a flange/bearing support structure. The bearings are preloaded by use of a spring. A stator sheet stack is wrapped with magnet wire and mounted to the bearing support structure whil e being electrically connected to a printed circuit board (PCB) , which also holds three Hall sensors that detect the main magnet position. maxon has the industrial heritage for the flat motor design, which has been built several million times, including over 1 million units for commercial truck exhaust gas cleaning systems , an application that provides real world proof of the robustness of the basic design. Flight Fidelity Modifications to the Catalog Designs It was known from previous maxon work that commercial motor designs usually require some modifications to allow reliable operation in a Martian environment , in particular , the wide temperature range cycling, Figure 3. maxon flat motor standard features

Document

AMS_2020.pdf

415 vibration/shock environment , and vacuum exposure. After deter mining that the Ø20 mm and Ø32 mm models provided the optimal power v ersus mass compromise, the commercial design was analyzed by a team incorporating both JPL and maxon personnel with the intention of applying the previous experience of both organizations to optimiz e the design for ultimate reliability in the harsh Martian environment . Table 2 lists the features in the COTS design that needed modification. Table 2. COTS design and modifications for Martian environment COTS Design Design Risk for Mars 2020 Application Solution Flange bearing carrier out of several parts with dissimilar materials Connection under shock/vibration; c orrosion ; coefficent of thermal expa nsion mismatch Create one- piece flange/bearing carrier Standard Hall sensors Temperature range limitations, failure due to radiation exposure Use Hall sensor that can be qualified for environmental conditions Rotor shaft glued to bearing races Shock and temperature environment breaks adhesive bond and causes rotor to move Add securing/spacer rings to design and mount all parts on shaft to hard stop. Weld rotor to shaft Main magnet only partially covered in adhesive Bond to rotor is not as strong as it could be and might fail under repeated temperature cycling Develop new bonding application process to ensure at least 80% coverage Flex-print PCB Can flex under vibration environment , which leads to mounted components breaking Replace wit h FR4 and flying lead wire connections Winding taps run directly to PCB ~1 cm lengths of unsupported thin wire susceptible to vibration- induced resonance and breakage Stake winding taps to stator and strain relief with “expansion loops” next to solder joint In order for the maxon catalog flat motor to survive the requirements of the Mars 2020 mission, several changes were made. The PCB material was changed from a flex print to FR4, a material commonly used for space missions as it does not significantly outgas, is robust enough to support the Hall sensors in a vibration environment , and fulfill s IPC-A-600 Class 3 requirements and testing required by JPL. The bearing preload architecture was modified so that the entire system was pushed to a hard stop meaning t he individual parts could not shift during impact or vibrations. A more robust spring was selected with shim ends to better distribute the forces and prevent the shim from breaching the bearing. High- quality industrial bearings were selected from one of maxon’s standard suppliers. Braycote 600EF was chosen as the lubrication with a fill-factor (15– 20%) of the free volume of each bearing measured (by mass) and documented. As has been standard practice for decades in space applications, the PCBs in both the M20 and M32 designs have conformal coating to: 1) protect the PCB from corrosion whil e in storage o n the ground awaiting launch; 2) protect against short circuits from any debris in weightless conditions ; and 3) suppress any tin whisker growth. maxon h ad a good experience working with Speciality Coating Systems in the Czech Republic for Parylene HT application for ExoMars and chose to work with them again for the Mars 2020 motors. The fully populated PCBs, including attached stators with windings were s hipped to Speciality Coating Systems , coated , and returned to maxon for integration into the rest of the motor. Undoubtedly the most difficult component choice to be made when designing a space- rated BLDC motor is the rotor angle detector required for th e control electronics to correctly commutate the motor. All of maxon’s BLDC motors use various types of Hall sensor s for this purpose. As described in Ref. [12], the solution for ExoMars was a radiation and temperature test program with many different types of compact Hall sensor s. Unfortunately, the Infinion TLE4945 sensor that was selected for ExoMars has been discontinued, but maxon had already tested and selected a variant of Honeywell’s bipolar SS41 H all sensor range for other space applications. JPL organized the purchase of a flight batch of sensors that were subjected to additional screening (pre- conditioning, lead tinning, and thermal cycling as well as lot qualification including Destructive Physical Analysis) to qualify the sensor lot prior to del ivery to maxon for population onto the PCBs. These sensors have been shown to operate well outside their published range, although performance deteriorates at the temperature extremes. Satisfactory operation can be achieved

Document

AMS_2020.pdf

416 between -100°C and +200°C. Radiation testing showed no failures after exposure to γ -radiation (absorbed doses of 300 J/kg for 64h). maxon’s flight heritage from the MER mission does not include gearboxes , however, maxon gained considerable experience during the ExoMars development [ 13] and qualification program. The program demonstrated that although COTS planetary gearboxes, in particular those developed for high- temperature “downhole” oil and gas industry applications, can be easily modified for space applications, significant lifetime problems arise when switching to vacuum- rated lubricants such as Braycote 601EF. ExoMars development work had shown that using a slurry of grease and oil as previously described for Harmonic Drive applications [14] partially mitigate d this lubricant issue. The root cause of the lifetime issues is that Braycote 601EF is fundamentally unsuitable for lubricating sliding (as opposed to rolling) surfaces. In order to achieve the lifetime requirements f or the ExoMars drill drive, the COTS design sliding motion between each planetary gear and its shaft/axis was converted to rolling motion by the addition of needle bearings , which was then used as the basis for the Mars 2020 gearbox (Figure 4) . Figure 4. Exploded view of M20 planetary gearbox stage Given the mission- critical nature of the M20 application, the JPL and maxon review teams decided that excessive residual risk remained in connections between sun gears and planetary carriers as well as between the planet axes pins and the carriers. In the COTS design, these connections are press fitted and welded, however , due to the hardened nature of the materials used for the axes and sun gears , the welds always exhibit ed surface cracks. This is an accepted compromise for industrial applications (and is not known to have ever caused a field failure), but due to the extreme Martian diurnal temperature cycling, this was considered to present an unacceptable risk of weld failure followed by axial movement of the components. The sun gear and planetary carriers’ connection was easily solved by combined these into a one-piece part (by using shaping manufacturing techniques). The problem of how to secure the axes was solved by adding a cage to the design. Both the ca ge and the planetary carrier have blind (vented) holes that the axes are pressed into, without requiring any further retaining mechanism. The cage is pressed onto an axial hard stop on the planetary carrier and is the n cold formed (swaged) into position. T he resultant design is then both weld- and adhesive- free and enables a free choice of gear and axes materials and hardening processes. No part involved more complexity to manufacture than the combined ring gear and motor flange for the M20 gearmotor. For industrial products , gearbox ring gears and motor flanges are usually screwed together on a large diameter thread and then either glued or welded. In the case of the M20 , however , the gearmotor needs to fit inside a ring of springs, which requires a star shape pattern of wires to be precisely oriented relative to the mounting lugs on the front of the gearbox. The disadvantage of the standard assembly method is that the large diameter thread does not allow a precise angular orientation of the gearbox to the motor. The solution to this problem was to combine the flange and the ring gear into one piece. This had

Document

AMS_2020.pdf

417 the additional advantage of allowing shorter and lower mass assembly and also eliminated any unreliability in the connection. The decision required a significant reordering of the assembly sequence, but this did not cause any further problems. T he manufacturing of the part itself, performed within maxon, involved a number of challenges to address, in particular th e deburring of the bottom of the ring gear teeth after shaping, which required an external supplier for an electrochemical machining step and another external supplier for sand blasting the many detailed features on the motor flange. Flight Design Issue s Investigated Detent Design Two versions of the M32 motor were designed, with different detent strengths, each sharing components except for the magnets generating the field. The axially charged magnet in the two designs had different lengths, creating a 10 mN •m detent and a 20 mN •m detent, with the larger length baselined to enable the same housing to be used for both motor types. The step size of 24 detents per revolution was chosen to correspond to the total commutation state changes of the motor design. When paired with a gearbox, the detent size produces holding torque at a small angle at the gearmotor output. The Hall sensors measure the edge field of the 4- pole pair continuous ring NdFeB magnet. The Hall sensors are on the output side of the motor for maxon’s flat series designs, which is the same side as the detent mechanism module. When the detent was added to the motor, the performance of the detent mechanism did not match the strength of the standalone unit and motor switching error was encountered. Measurement of the magnetic field revealed the single axial magnet had significant stray flux in the vicinity of the Hall sensor, causing the switching error and reduced detent strength ( Figure 5). Figure 5. Left : Compensation magnet measurements for stray flux ; Right : Computed T omography ( CT) scan cross section of final assembly Two corrections were made. First, a compensation magnet axially charged in the opposite direction was added to the detent magnet stack, strengthening the field through the detent wheel. Second, the axial gap between the edge of the rotor cap ring magnet and the Hall sensor was reduced by shimming each motor individually after measuring subassembly dimensions. The change in axial gap increased the magnetic field strength and hence the signal -to-noise ratio in the H all sensor.

Document

AMS_2020.pdf

418 Low-Quantity Grease Dosing For the M20 gearmotor, JPL required r apid acceleration to the commanded speed without an excessive current draw over the entire specified t emperature range. This is a problem due to the high viscosity of Braycote at low temperatures. The standard quantities of grease used per gearbox stage by maxon for COTS products are determined by the requirement to deliver the longest possible lifetime wi th standard environmental conditions. The use of Braycote required the standard quantities to be reassessed. A series of tests were performed with gearboxes built using differing quantities of grease. In each case, high frequency resolution data of the start -up current draw was collected. It was necessary to reduce the grease load by a factor of 10 compared to the standard dose in order to obtain acceptable start -up performance. Subsequent life testing demonstrated that the require d life (low compared to industrial applications) could still be achieved. For the flight units , application using the grease plating method was used to ensure the low quantity of grease was uniformly distributed, a process that is unnecessary for industrial applications, where only manual dosing with a syringe at defined locations is sufficient. The mechanism driven by the M20 actuator has a limited linear stroke. The cold drag performance of the actuator had significant transience in the first 5 seconds of motion; this large variation in torque production was outside the system- level requirements for controllability. JPL tested three actuators at several low temperatures in order to characterize the start -up current as a function of temperature. Figure 6 shows a knee in the drag happening between -30°C and -40°C for this gearbox . The actuator still exhibited margined torque production at the - 70°C operation temperature limit, but due to controllability and other friction effects in the mechanism at - 70°C, the heater was modified to increase the low -end operating temperature. Figure 6. M20 drag current temperature dependence Shock and Motor Bearings The M32 motor was tested (along with the mating gearboxes) to withstand a pyroshock environment simulated by a tunable beam. The pyroshock spectrum had a peak level at 3,000- g with a knee frequency at 1,600 Hz. After testing, the motor operated with a significant rattle, a noise equated to that of a coffee grinder. Although CT scans did not reveal gross damage in the assembly, sever al other non- destructive tests, including a noise frequency analysis and rotating shaft deflection test, indicated motor bearings as the damaged component. The frequency analysis had peaks at the motor bearing inner ring frequency that would scale with mot or speed and the deflection test had a signature reminiscent of dimples in the raceway allowing up to 0.0127 mm of deflection. The motor bearings were disassembled and inspected. The motors ’ front bearing inner ring, outer ring, and balls were deformed in both directions from the tunable beam test (Figure 7). The back bearing did not have any damage.

Document

AMS_2020.pdf

419 Figure 7. Damaged motor bearings from shock event The g- level induced by pyroshock was initially analyzed with respect to the brittle components and solder/ strain relief only, and did not consider the spring/mass system of the bearing/rotor combination. A design change between COTS and flight had the front bearing hard mounted in both directions, allowing the back bearing to float, resulting in the front bear ing absorbing all of the energy of the accelerated rotor. The M32 motors are used in multiple locations across the Mars 2020 rover and any changes would cause a significant project schedule impact. In most locations, the pyroshock requirement was over specified; the hardware was requalification tested to the new pyroshock requirements. The motor, and subsequent gearmotor tests, all passed the new pyroshock exposure and long rotational life requirements. There remained a single location where the motor had exposure to the high pyroshock environment but with a short rotational life. With the project considering the risk, the solution was to use the damaged motor to qualify a deployable device with high shock and limited motor rotational life. The qualification demonstrated that the damaged motor could achieve the limited revolutions needed. The flight unit was delivered and the mechanism underwent qualification at the next level of assembly. The pyrotechnic deployment was demonstrated four times and did not pr oduce a “coffee- grinder” motor, showing that the either the requirement or tunable beam test was over conservative. In both solution cases, no changes to the flight motor were made. Overheat ing Concern The coil -to-coil resistance for the M32 and M20 motor s are 14 ohms and 28 ohms , respectively. Additionally, the copper wire is wound around two PEEK stator caps at each end of the stator lamination stack, which provides thermal isolation from the bulk stator mass. The current driving the motor causes significant Joule heating due to the high resistance, quickly increas ing the temperature of the low mass of copper wire insulated from the stator, which can lead to thermal runaway. Th e overheating worsens as the current reaches 1A. Due to design principle requirements to demonstrate 2× torque at operational temperature limits, hot dynamom eter tests are the highest risk activities to perform on the ground. The M32 has an additional thermal risk due to the detent mechanism; the larger holding torque design was roughly equal to the nominal performance of the motor, so twice as much current is needed during slow speeds (where the rotor has no signifcant inertia to carry it over the detent peaks) compared to fast speeds. A thermally instrumented unit was used to create a thermal model to understand this interaction. Guidelines for operation early on in the Mars 2020 design and qualification program were related to current input and time —a current profile could be estimated, the heat input was analyzed to confirm no overheat condition was achieved, and then a wait period was enforced in order to ensure the test article reached steady state.

Document

AMS_2020.pdf

420 The mechanism testing and subsequent system- level testing had sequences and current profiles that could not be predicted. The solution was to use a simplified thermal model on both the rover and control electrical ground support equipment (EGSE) , which was the same implementation used on the MER rovers. The model is supplied with a case temperat ure from a platinum resistance thermometer ( PRT) and the current telemetry feeds the Joule heating. The thermal protection watches for nodes and can shut down motion if temperatures r ise above certain limits. Acceptance and Qualification Testing An acceptance test program was implemented to determine JPL acceptability of the M32 and M20 motors. The acceptance testing centered on workmanship and functional performance of each unit , whereas a qualification program subjects designated units to the gambit of JPL- margined dynamic and thermal environments and life tests to verify that the design of units from the same production and manufacturing lots would meet the mission requirements. Qualification was completed at the gearmotor level, with some actuators further qualified at the mechanism level. Portions of the M20 gearmotor qualification testing were completed at maxon using their testing facilities, where feasible. The residual tests were completed at JPL due to infrastructure availability as well as long -duration (life test) activities outside of the scope of maxon’s contractual effort. All M32 FM units were subjected to the acceptance tests listed in Table 3 , with defined criteria when appropriate. No-load characterization was used as a functional check to confirm the specific motor was performing as expected and within family for the M32 flight motors. Table 3. M32 motor acceptance tests Notes: CW = clockwise; CCW = counter -clockwise The M20 motor was subjected to similar acceptance testing (Table 4) prior to integrati on into the M20 gearmotor . M32 Motor Acceptance Test Test Conditions Acceptance Criteria Dielectric strength 500V AC, 60 Hz for 60 ±5 seconds I < 1 mA Insulation resistance (R) 500V DC, for 60 ±5 seconds R > 100 MΩ Electrical bonding N/A R (flange to housing) < 25 mΩ R (flange to shaft) < 100 Ω Run-in 60 seconds CW and 60 seconds CCW at 28V N/A Hall effect Integrated NonLinearity (INL) sensor Maxon defined test <10° INL (target <7.5° ) Natural cogging torque (T) harmonic magnitude CW and CCW over two revolutions, speed 2 rpm T = Detent Value (10 or 20 mN•m) ± 10% Powered torque ripple plot Powered in direction of motion, powered opposite direction of motion N/A (characterization only) , validates torque is produced at all rotor positions Load 12V/20V/28V, CW and CCW, ambient temperature Motor torque constant = 50 mN •m/A ± 10% Start -up sensitivity CW only, 22°C /+70°C /-70°C N/A (characterization only) No load 28V, CW only, 22°C /+70°C /-70°C /22°C N/A (characterization only) “Launch” random vibration Frequency (Hz) Protoflight/Design Verification (Qual/PF) Test Level 20 0.15 g2/Hz 30–90 0.5 g2/Hz 166–450 0.08 g2/Hz 2,000 0.0040 g2/Hz Grms 10 (in all 3 axes, 1 min each) N/A (exposure only)

Document

AMS_2020.pdf

421 Table 4. M20 motor acceptance tests The acceptance tests listed in Table 5 were c ompleted at the M20 gearmotor level, which included thermal environment exposure at non- operational temperature limits of - 135°C. Table 5. M20 gearmotor acceptance tests The majority of the M32 and M20 hardware subjected to acceptance testing met performance requirements and were accepted as meeting flight fidelity. JPL accepted the risk of qualifying the M20 actuator post delivery of the flight units due to the schedule criticality of needing the flight hardware assembly. Tests completed by JPL to fully qualify the design (Table 6) exposed an actua tor to the flight acceptance test (with the exception of 2 minutes per axis random vibration exposure) and then completed a long-duration thermal bakeout, pyroshock , and humidity exposure. The gearmotors underwent a long- duration thermal vacuum bakeout to replicate the maximum expected exposure the assembled mechanisms may be subjected to in order to meet Planetary Protection requirements. The gearmotors were also subjected to 40 hours of accelerated corrosion exposure in an 80°C/80% relative humidity environment meant to simulate any corrosion that may occur in the gearbox while exposed to the Earth’s atmosphere before launch. The gearmotor was then subject to the rotary life test. A separate motor -only unit was subject ed to the thermal life test . M20 motor Acceptance Test Test Conditions Acceptance Criteria Dielectric strength 500V AC, 60Hz for 60 ±5 seconds I < 1 mA Insulation resistance 500V DC, for 60 ±5 seconds R > 100 MΩ Electrical bonding N/A R (flange to housing) < 25 mΩ; R (flange to shaft) < 100 Ω Run-in 60 sec CW and 60 sec CCW at 28V N/A Hall effect (INL) sensor Maxon defined test <10° INL Torque ripple CW and CCW over one revolution N/A (characterization only) , validates torque is produced at all rotor positions No load 24V, CW and CCW, 22 °C N/A (characterization only) M20 Gearmotor Acceptance Test Test Conditions Acceptance Criteria Run-in 7.5 min utes CW and 7.5 min utes CCW at 24V N/A Back drive torque Ramp to 75 mN •m, CW only Static holding torque: > 50 mN•m No load 24V, CW only, STP N/A (characterization only) “Launch” random vibration Frequency (Hz) Protoflight/Design Verification (Qual/PF) Test Level 20 0.15 g2/Hz 30–90 0.5 g2/Hz 166–450 0.08 g2/Hz 2,000 0.0040 g2/Hz Grms 10 (in all 3 axes, 1 min each) N/A (exposure only) Non-operational thermal One thermal cycle from +113°C to -135°C Two thermal cycles from +70 °C to -135°C Restrain static holding torque (T > 50 mN•m) Operational thermal Start -up sensitivity No load Load Each test performed at +70°C /22°C /-70°C CW only 24V, CW only 12V/20V/28V, CW and CCW Gearmotor torque constant = 1620 N •m/A ± 10%

Document

AMS_2020.pdf

422 The rotary life qualification unit completed its life test with minimal wear, and the design completed qualification for flight as of June 2018. The Thermal Life test is 1.5 years in duration and was successfully completed in December 2019. Table 6. M20 gearmotor qualification tests. M20 Gearmotor Qualification Test Test Conditions Acceptance Criteria Thermal vacuum bake -out 113°C ±2°C at 1.3 × 10-3 Pa for 288 hours N/A (exposure only) Pyroshock Tunable beam (2 hits in 2 axes) Frequency (Hz) Shock Response Spectrum ( SRS) Level (Q = 10) 100 5 g 100–3,500 +9.0 dB/Oct. 3,500 –10,000 1,000 g N/A (exposure only) , in family with no- load characterization Humidity exposure 40 hours of 80% humidity/+80 °C N/A (exposure only) Rotary life test Torque bins: 0.2 N •m for 40,400 revolutions 0.3 N •m for 12,400 revolutions 0.4 N •m for 2,100 revolutions 0.5 N •m for 1,500 revolutions Split between CW and CCW and alternating between +70°C, +22°C, -55°C , and -70°C In fami ly with no- load characterization Thermal life test 3015 thermal cycles (4.5 Mars years) broken into seasonal profiles ( -80°C to +85°C for summer and - 115°C to +50°C for winter) In family with no- load characterization at interval checks Final Delivery For the nine flight locations of the M32 motor across the Mars 2020 rover , a total of 37 gearmotors were needed to meet flight, flight spare, engineering model (EM) , and qualification units. m axon required a minimal build to efficiently use their production and manufacturing series, which resulted in three lot builds (EM1, FM1, FM2) totaling 99 delivered M32 units. The 99 units consisted of 10 EM1s (3× 10 mN •m and 7× 20 mN•m), 35 FM1s (14× 10 mN•m and 24 × 20 mN•m), and 54 FM2 s (33× 10 mNm and 21× 20 mN•m). Similarly, the M20 gearmotor had a JPL need of 6 units but a total of 40 units were delivered. The 40 units consisted of 16 EMs and 24 FMs. On the FM units, the carrier s with sun gear were made out of two different material t ypes, with 12 units of each type delivered. An attrition of over 50% during the manufacturing and production phases had delivery effects but the flexibility and availability of the on- site facilities allowed maxon to recoup schedule as best as possible and phase deliveries between the lots as needed. Due to the surplus of flight units available, JPL was able to cherry -pick motors with the optimum performance and flight fidelity for the 10 flight units. The residual units are logged in the JPL flight hardware logistics program for future NASA project us e. Additionally, maxon has a catalog flight -qualified specification for the M32 and M20 flat motors.

Document

AMS_2020.pdf

423 Conclusions and Lessons Learned Detent Design Although the detent brake worked well and the add- on module that was designed proved capable of passing all qualification tests, this solution adds mass and complexity. In parallel, work was carried out by maxon to make rotors with segmented magnets (rather than a one- piece ring magnet) and then shape the magnetic field from those magnets to create the necessary detent. Making magnetic field simulations that correctly predicted the obtained detent was difficult, which resulted in initial prototypes deliveri ng insufficient torque. Follow -up prototypes achieved the targeted 10 mNm. The greatest difficulty encountered was fixing the individual magnets to the rotor so that they were not susceptible to breaking under shock impacts. Eventually it was decided that this presented too much risk, especially to the schedule if several attempts were needed to find a workable solution. Despite not being chosen for the flight design, with further development this is likely to be a more reliable (because of fewer parts) and compact solution that should be investigated further. Overheat ing Concern The overheating concern has led to significant effort s in both analysis and implementation to reduce the motor wire temperature below the allowable limits. For past missions, prot ecting the motors from overheating entailed installing a PRT on the case and operat ing under the assumption that the housing temperature is well coupled to the wire temperature. Due to the outer rotating design of the Mars 2020 motors , there is no external location that is thermally coupled. However, a potential solution for both ground testing and thermal model validation would be to implement a thermocouple (TC) directly into the windings at the time of assembly and provide a path for those thermocouples to be moni tored, either through traces in the board or hol es in the board that can be routed to the channel for the lead wires. These ther mocouples can be set to a GSE thermal alarm or act like a switch. The TCs can monitor concurrently while running the thermal model in multiple atmospheric environments to better ground the model in demonstrated performance. An additional improvement to the design could be to thermally couple the wires to the stator lamination stack instead of suspended between t wo PEEK en d caps. Teamwork between a Government -Funded Science Institut ion and a Commercial Company A close collaboration between maxon and JPL was essential in developing the flight units, as it required a combination of maxon’s decades of experience with industrial motor and gearbox design and JPL’s decades of institutional experience in spacecraft design. For such a collaboration to work, it is necessary for both sides to understand each other’s motivation s and methodologies . Commercial companies (especi ally those with limited previous space experience) need to understand that organizations like JPL can (and do) place significant resources into issues that may be of no importance in an industrial setting, which is a result of decades of hard learnt lessons where a single small failure can terminate a mission. The space environment is very unforgiving of design or manufacturing errors , which may never be noticed in a standard Earth-based application. The education of contractor staff on this history is a cr itical part of gaining acceptance for the space science mission method of working. In turn, government -funded institutions need to understand how industrial manufacturing companies , such as maxon, usually have the goal of using their limited engineering resources as efficiently as possible to keep the production lines filled with products that receive repeat orders. Thus, one-off high engineering resource developments are often viewed as a disturbance to the series production goals of the company ; however , this can be partially mitigated by presenting the work as subsidized technology development and being as open as possible in helping the commercial company obtain a marketing effect by publicizing the work. Final Thoughts During the three- year collaboration, significant resources were required from both organizations to resolve the problems encountered and complexity involved in developing the Mars 2020 actuators . Even when the basis of the design was a commercial product, the issues encountered highlight the importance of allowing

Document

AMS_2020.pdf

424 sufficient time in mission planning for solution develop ment . It was pivotal for JPL and maxon to leverage experts at each location to address critical problems (along with using a 9- hour time difference between Luzern and Los Angeles to accomplish an 18- hour project work day). Both the M20 and M32 designs have been successfully qualified. The M32 designs were integrated into three different, separately sourced gearboxes and passed all qualification testing. The two versions of the M32 motor are implemented into nine locations on the Mars 2020 rover and the M20 is integrated into a single location. Flight model hardware has been built, acceptance tested, integrated into the target gearboxes and mechanisms , and installed on the Mars 2020 rover awaiting launch to Mars in July 2020 (Figure 8). Figure 8. maxon hardware installed on Mars 2020 Acknowledgements The successful collaboration and delivery of the maxon flight motors to Mars 2020 would not be possible without many individuals at JPL and maxon over the three- year effort. In addition, the authors would like to thank their families who endured years of middle -of-the-night teleconferences and nights away on international trips. The research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration.

Document

AMS_2020.pdf

425 References 1. NASA Procedural Requirements , NPR 8705.4 (2004), https://nodis3.gsfc.nasa.gov/npg_img/N_PR_870 5_0004_/N_PR_8705_0004_.pdf . 2. Braun, David and Don Noon, “Long Life” DC brush motor for use on the Mars Surveyor Program, 32nd Aerospace Mechanisms Symposium (1998). 3. See graphic on page 2 of Ref. [5 ], or this direct link: https://mars.nasa.gov/imgs/2013/07/mars - exploration- timeline- 07-2016 -hpfeat2.png . 4 Office of Inspector General, NASA’s Mars 2020 Project, Report no. IG -17-009 (2017) , https://oig.nasa.gov/docs/IG -17-009.pdf . 5. NASA, Mars 2020 Mission Overview , available at: https://mars.nasa.gov/mars2020/mission/overview/ . 6. Szwarc, Timothy, Jonathan Parker, and Johannes Kreuser, STIG: A Two- Speed Transmission Aboard the Mars 2020 Coring Drill, 45th Aerospace Mechanisms Symposium (2020). 7. Chrystal, Kyle, Percussion Mechanism for the Mars 2020 Coring Drill, 45th Aerospace Mechanisms Symposium (2020). 8. Barletta, Anthony, Design and Development of a Robust Chuck Mechanism for the Mars 2020 Coring Drill, 45th Aerospace Mechanisms Symposium (2020). 9. Silverman, Milo and Justin Lin, Mars 2020 Rover Adaptive Caching Assembly: So Many Challenges, 45th Aerospace Mechanisms Symposium (2020). 10. Grimes -York, Jesse and Ken Glazebrook, Sealing Station Mechanisms for the Mars 2020 Rover Sample Caching Subsystem, 45th Aerospace Mechanisms Symposium (2020). 11. Brown, Adrian, Mars Science Laboratory: the technical reasons behind its delay , The Space Science Review (2009), http://www.thespacereview.com/article/1319/1 . 12. Phillips, Robin R., M. Palladino, and C. Courtois , Development of Brushed and Brushless DC Motors for use in the ExoMars Drilling and Sampling Mechanism , Aerospace Mechanisms Symposium (2012). 13. McCoubrey, R., et al., C anada’s Suspension and Locomotion Subsystem for ExoMars 2 018, 65th International Astronautical Congress , Toronto, Canada (2014) . 14. Herman, Jason and Kiel Davis , Evaluation of Perflour opolyether Lubricant Lifetime in the High Stress and High Stress -Cycle Regime for Mars Applications , 39th Aerospace Mechanisms Symposium (2008) .

Document

AMS_2020.pdf

427 Mars 2020 Center Differential Pivot Restraint : Flexurized Spring System Providing Compliance for Rover Mobility Deployment Prior to Landing Matthew Cameron* and Kevin Liu* Abstract The Center Differential Pivot Restraint (CDPR) for the Mars 2020 mission is designed to constrain motion of the rover differential and to dampen out the load response during rover’s mobility deployment event in Entry, Descent, and Landing (EDL) phase of the mission . For Mars 2020, it was required to redesign the mechanism to save minimum 33% of the mass from the previous Mars Science Laboratory (MSL) design . The new mechanism architecture used a monolithic spring- like flexure to act as the system compliance, resulting in 50% mass savings from the MSL design. Along the design and prototype testing process, several mechanism design principles were re -affirmed as well as lessons learned about potential race conditions. Introduction The Mars 2020 mi ssion is designed to land and traverse a newly improved rover on Mars to gather soil samples that could be returned to Earth for detailed study. During the EDL portion of the mission, the rover is lowered from the sky crane via three bridle connections while the rover’s mobility subsystem’s rocker/bogie suspension on both side s of the rover deploys in preparation for surface touchdown (Fig. 1). During this deployment event, the CDPR absorbs and dampens out the induced load into the rover at the chassis top deck interface in the event when the side- to-side suspension deployment becomes asymmetric. Same as that of the MSL mission, the CDPR is mounted on the top deck of the rover chassis and located directly aft of the rover differential assembly (Fig. 2). As improved upon the previous MSL CDPR [1], completely new design and architecture for the Mars 2020 CDPR were achieved by maintaining the same functionalit y, while reducing the restraint’s mass by 50%. From conceiving the initial design in November 2015, through multiple design reviews and prototype testing, to completing flight integration and qualification testing around early January 2020, the entire crad le-tograve process spanned over a 39- month period. This paper discusses details of the new Mars 2020 design and mechanism features, its development process, key improvements from the previous MSL mission, and the design validation testing campaign; highli ghting key lessons learned throughout the design and testing process. Mars 2020 vs. MSL Design The design for Mars 2020 CDPR is composed of a spring- like flexure, constrained by a pyro pin- puller attached to the mobility differential on one end and a shoulder bolt to a bracket bolted to the rover chassis top deck on the other end (Fig. 2 ). Details o f its design and the deployable mechanism are presented in a later section. For brevity and scope of this paper, see [1] for complete details of MSL’s CDPR design. In both designs, the mechanism’s function is the same and is described in three phases: (1) rover mobility deployment response dampening and motion constrain during EDL, (2) CDPR mechanism deployment, and (3) remaining deployed for the remainder of surface operations. Major similarities of CDPR between the two missions include how each deployabl e is constrained to the differential as well as the main passive element in the deployment mechanism. * Jet Propulsion Laboratory, California Institute of Technology, Pasadena, CA Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

428 Figure 1. Mars 2020 EDL timeline indicating when the CDPR is actively engaged during the m obility deployment just prior to touchdown. Figure 2: MSL CDPR (left) v ersus Mars 2020 CDPR (right)

Document

AMS_2020.pdf

429 The deployable for MSL and Mars 2020 are both constrained in place by a 9.5- mm ( 3/8-inch) diameter pinpuller pyro device. The pin-puller is in the same location relative to the Center Differential Pivot ( CDP ) and differential (Fig. 2). The MSL’s CDPR dampened out the response from the asymm etric load using two mirrored stacks of b elleville spring washers within a crank -slider mechanism [1]. A titanium linkage (weigh ing approximately 150 grams) coupled the pin -puller to the belleville spring washers . Immediately after the pin- puller was fired, only the titanium linkage was deployed and constrained in place by two torsion springs [1]. For Mars 2020, a steel spring- like flexure replaces the titanium linkage and the entire belleville spring assembly. Immediately after the pin-puller is fired , the steel flexure ( weigh ing approximately 1500 grams) is deployed to a final stowed position and is also driven by two torsion springs . This is where the similarities between MSL and Mars 2020 CDPRs end. Since the mass of the deployable for Mars 2020 was 10 times greater than that of MSL , a heavy emphasis on CDPR’s ability to deploy its flexure and to ensure it remains deployed was put forth during the re- design process. As a result, t his drove to four major mechanis m design changes from MSL to Mars 2020. A summary of these changes are as follows : • Size and strength increase of the torsion springs in order to achieve the required release torque • Installment of a kickoff spring as a tertiary flexure release aid • Installment of a magnetic latch into the deployable hard stop to further ensure the flexure to remain at a fully and final deployed position • Having the spring element being between the spherical bearings ensured moment decoupl e or pure axial load transfer which eliminate d alignment concerns that lead to several issues in the MSL’s mechanism [1] Details of the aforementioned mechanisms and structural features are discussed in later sections . Lastly, even though the location of the CDPR relative to the differential was similar between MSL and Mars 2020, the mechanical interface between the rover chassis top deck had to be redeveloped. Part of what drove the MSL design was the shear pin interfac e between the CDPR and the top deck that not only required a time-consuming match drilling process but was also set by the time the compliant MSL CDPR had to be developed [1] . For the Mars 2020 design, this interface was renegotiated to avoid using match drilled shear pins and allowed for easier integration to the chassis top deck. Instead, a singular shear feature was created. The shear feature is a monolithic planar extrusion with a t ightly toleranced diameter from the bottom of the chassis bracket that mated to a pocket on the rover chassis top deck. Along with the four bracket mounting fasteners, the designed size and fitment tolerance of the shear boss were intended to take the enti re shear load through the CDPR if the frictional cap ability of the preloaded interfaces common to the four bolts was compromised . This newly designed shear feature ensured full shear capability and eliminated any risk of joint interface slippage. This design feature allowed for easy integration and avoided a high risk and precise match drilling operation on the major top deck component. This was a good reminder of a design engineering principle to value design for assembl y. The interface between the CDPR and the mobility differential utilized the same match drilled shear pins and interface as MSL’s design due to mobility differential remaining nearly heritage from MSL. Despite this particular interface still being match dr illed, the re- designed shear boss feature of chassis bracket to top deck interface still required less time and effort. Flexure Conceptualization and Trade -off for Mars 2020 As with MSL, the structural and compliant aspect for the Mars 2020 CDPR mechanism was to create a theoretical torsi on spring about the CDP [1] . Unique for the Mars 2020 CPDR was to accomplish this with a lower torsional spring stiffness and overall assembly mass i nstead. As determined by simulation of EDL sequence in ADAMS , a reduced torsional stiffness about the CDP coupled with a reduced CDPR mass would linearly reduce the peak flight loads transferred into the rover chassis and mobility system. This finding was set forth as the design objective for the Mars 2020 CDPR where MSL’s torsional stiffness of 200,000 N -m/rad [1] was halved to 100,000 N -m/rad to become the new Mars 2020 design parameter for the CDPR .

Document

AMS_2020.pdf

430 The reduction in effective torsional stiffness and mass reduction did come with a trade- off, however . The linear spring of the CDPR would be less stiff , yet still had the same amount of time to dampen out the resultant load at the center of the differential . While the load to dampen out from an asymmetric deployment would start -off lowe r, the more compliant linear spring would not be able to dampen out as much load as the MSL CDPR. This resulted in a higher load at the time of pin-puller firing for the Mars 2020 compared to MSL. From worst case ADAMS models with the softer Mars 2020 restraint system, the peak load (torque about the CDP axis or rover z -axis as denoted by “Mz”) at the time of release was roughly 56% higher, but still within the capability of the pin- puller (Fig. 3). Figure 3. Moment about CDP of MSL vs. Mars 2020 showing the maximum torque through the CDP at the time of CDPR pin- puller firing: (a) MSL loads showing an initial higher peak mobility deployment load and a lower pin- puller firing load, (b) Mars 2020 loa ds showing an initial lower peak mobility deployment load and a higher pin- puller firing load. Flexure Design and Development With the target torsional stiffness about the CDP set, the design and placement of the spring- like flexure that would act as the damper for the CDPR was the first component of the CDPR to be fully prototyped and tested. The flexure design began with a simple cantilever beam concept. The multi-looped feature was then implemented and modeled by placing several cantilever beams in series to form a linear spring. In the early design phase, a MATLAB optimization program was created to determine the placement and sizing of the flexure in order to minimize mass while also achieving the target equivalent tor sional stiffness about the CDP. A structurally optimum and mass efficient location of the flexure was determined to be the same as that of MSL; having the pin- puller attaching to the flexure 180.5 mm aft of the CDP axis. After this optimization determining rough sizing of flexure dimensions, the design transitioned to higher fidelity Finite Element Analysis (FEA) modeling to close in on a more final design of the flexure. There were challenges in the initial design of the flexure to strike a balance between required material strength and structural buckling margin as well as target flexure compliance. To begin, Maraging 300 steel was selected as candidate material for the flexure due to its superior strength properties and high strength vs. stiffness ratio as material strength was the first design constraint to satisfy followed by structural compliance. To arrive to the first prototype design, extensive analytical iterations via FEA were performed in order to meet a rather narrow design window of stress, buckling, and stiffness requirements . Findings from the structural analysis indicated that peak stress magnitudes and distribution as well as overall flexure stiffness are highly sensitive to the width of the flexure loops followed by stem of the loops (Fig. 6). Prior to Detailed Design R eview (DDR) , a flexure prototype out of the flight material specification and same lot of flight material was subjected to a series of structural characterization testing. Objectives of this testing was to statically characterize the flexure’s uniaxial tension and compression stiffness at room temperature and qualification level hot (+40°C) and cold ( -55°C) temperature up to CDPR’s flight limit load (FLL) . A final test would be conducted to determine the flexure’s tensile residual strength capability at room temperature.

Document

AMS_2020.pdf

431 Since flexure strength and stiffness were two self -competing design constraints and overall CDPR compliancy in mobility deployment load absorption is entirely depending on the flexure stiffness, its physical measurement would serve invaluable to confirm FEA predictions. Fig. 4 show s the images of the nominal test set -up as well as deformed shapes under both tension and compression. Figure 4. Fully reversed range of deflection of the p rototype flexure at (a) max extension from tensile load, (b) nominal unloaded state, and (c) max compression from compressive load. Below are the key findings from this flexure prototype characterization test: • Flexure’s uniaxial stiffness (load vs. deflection) appeared highly linear and repeatable at both room (illustrated in Fig. 5 ), hot, and cold temperatures with little to no signs of hysteresis over a set of 34 tension- compression cycles • Temperature effects are minimal – within a maximum difference of 2% of the measured flexure’s stiffness relative to the desired target stiffness • Residual strength test achieved a tension load of approximately 46 kN, which demonstrated a 33% more tension load capability than predicted by linear FEA • No signs of material yielding observed throughout the entirety of the test This prototype testing sufficiently demonstrated that flexure critical stress concentrations exhibited morethan- adequate strength capability. Based on this, nonlinearity in material’s elastic -plastic behavior as well as large displacement theory were incorporated in the FEA. This updated the predicted onset of yielding in the flexure to be approximately 71 kN, which was more than twice the initial predicted ultimate failure load, and 40% higher than the max load reached during the residual strength charact erization test. This warranted the final round of flexure thinning to further reduce its stiffness and mass (Fig. 6).

Document

AMS_2020.pdf

432 Figure 5. 15 overlaid cycles of raw force vs. displacement data at room temperature. Figure 6. The evolution of the flexure design: (a) pre- Initial Design Review design, (b) proposed Detailed Design Review design, and (c) final delivered flight design. The trade- off being after this final round of flexure thinning out , was that the flexure’s lowest structural margin became buckling critical rather than based on local stress concentrations. Buckling test of the flexure was subsequentially performed post -DDR on a flight like design. Test results indicated no signs of structural buckling up to the required 1.4 test factor over the FLL presented at DDR. Resulting from this f lexure developmental design and testing, the final design was proposed at DDR and was hardly changed for the final delivered flight flexure configuration post-DDR. Deployment Mechanism This new architecture of combining the compliant component (the flexure) and the deployable resulted in a redesign of the elements of the mechanism to deploy the flexure. More often than not, it required increasing size as well as strength of these elements . While the Mars 2020 CDPR had more components, the different approach of the mechanism resulted in approximately 50% less mass for the whole assembly while actuating a 10 times more massive deployable. The MSL deployment mechanism used symmetrical torsion spring to deploy a 150-gram linkage. The Mars 2020 deploy s the 1500- gram flexure using larger asymmetric torsion springs, initial motion kickoff s pring, and a magnetic latching hardstop to contain it in its deployed state. This section details the development of these elements of mechanism in the Mars 2020 CDPR as well as several findings and key lessons learned throughout the development process.

Document

AMS_2020.pdf

433 Torsion Springs The main components to deploy the fl exure are the torsions springs. At the time of DDR, it was required to show by design twice the needed torque to deploy the flexure to overcome mechanism frictional losses in dynamic environments. For Mars 2020, each torsion spring increased in size and strength compared to the MSL CDPR torsion springs . These larger torsion springs led to a design challenge of how to positively capture the two free ends of these larger torsion springs. For MSL, the torsion spring ends were manually bent to route the spring wire end through a hole in the t itanium linkage . For larger Mars 2020 torsion springs, the spring wire was fabricated out of a stronger stainless -steel material and from spring wire nearly twice the diameter . There was concern if one could manually bend the spring wire accurately and without damaging the spring wire. Thus , it was chosen to have the vendor make the torsion spring with prescribed bends at each end. The style of bends at the end of the torsion spring to route through the flexure was inspired from linkage rods found in automotive engines. There are t wo 90° bends at one end of the torsion spring and are route d through a hole in flexure. Then once the torsion spring is installed on its arbor, each torsion spring would be positively captured. The size and position of the pass -through holes in the f lexure and the torsion spring bends were designed together to achieve this positive retention of torsion spring end in the flexure. In the end, this design proved to be easy to implement and less susceptible to damaging the hardware with a manual bending process. See Fig. 7 for photos of wire end bends of each torsion spring. Figure 7. Views of torsion spring ends: (a) upper torsion spring, (b) lower torsion spring and the travel limiters are protruding through the chassis bracket clevis with set gaps to the back shelf on the flexure. There were design space limitations on the lower of the two torsion springs, which resulted in an asymmetrical design. As a torsion spring is preloaded, the effective spring diameter is decreased and the overall spring length is increased. The lower torsion spring had a fixed height from the confined space between the chassis bracket bottom and top deck. There was concern that the lower torsion spring could get to o long when the flexure is in its stowed state. The upper torsion spring did not have the same height limitations and could have more loops and delivery more torque for the same deflection. The combination of the two asymmetric springs surpassed the DDR requirement of 100% torque margin, roughly 66% of the combined delivered torque came from the upper torsion spring. However, the fault case of only the lower torsion spring acting was enough still to deploy the flexure. To verify this release redundancy capability, various prototype and Dynamic Test Model (DTM) tests have indeed demonstrated that the mechanism could still deploy fully without the upper, stronger, torsion spring contributing to the deployment.

Document

AMS_2020.pdf

434 Initial Motion Kickoff Spring To adhere to f light mechanism design principle s, a kickoff spring to aid in the initial deployment of the flexure was required to be added after DDR. This leaf spring, also made out of corrosion resistant steel ( CRES) , was placed within the c levis gap of the CDPR’s bracket connected to the aft of the differential , the Horizontal Swing Arm (HSA) bracket. With its preload, t he spring was designed to exert a force on the flexure lug to minimally displace the flexure approximately 9.5 mm, or the diameter of the p in-puller pin. Prototype testing yielded two find ings: first that the leaf spring did not need conditioning, b ut that the end interfacing with the flexure was digging into the paint of the flexure and causing the paint to shed. Tw o actions were taken from this lesson learned: (1) incorporate a slight curvature on the kickoff spring end contacting the flexure for a smooth point of contact, and (2) mask the potential contact area of the kickoff spring and flexure lug side to avoid any flexure paint scraping. Lastly, a coating of dry -film lubricate was applied to minimize friction of the contact surface on the kickoff spring. See Fig. 8 for the kickoff spring configuration with the flexure in the stowed position. The addition of this spring had several benefits for designing a flight mechanism. Not only did it add extra deployment force for the mechanism, but also fail-safe capability in a fault case to ensure the mechanism functions as intended if the larger and more capable upper torsion spring functionally or structurally fails . This more robust fault case configuration significantly boosted confidence that the CDPR will work as required for the critical separation prior to touchdown on the Martian surface. All these benefits are an important lesson learned of why it is flight design principle to have an initial motion spring for a deployable. Figure 8. Views of initial kickoff spring: (a) view of kickoff spring attached to the HSA bracket, (b) view inside the bracket clevis where the spring interfaces with the bare flexure lug. Travel Limiters With the new Mars 2020 design, t he larger height and length of the flexure as well as the rotational degrees of freedom in the spherical bearings at two ends of the flexure introduced a few close clearance concerns in the mechanism. At nominal (when flexure is engaged or not yet released), the 10.3 mm vertical clearance between the chassis top deck and flexure was found insufficient in the event if the flexure pitches downward when it exits out of the HSA bracket during deployment. During testing without the torsional springs, there was a roll rotational degree of freedom in the flexure, which allowed flexure contact with the chassis bracket due to residual rotations from the spherical bearings at two ends of the flexure. To address these concerns of interference, features on the chassis bracket were implemented to adapt four (4) vertical travel limiters at the clevis region of the chassis bracket . Two of them are opposite of and vertically aligned with each other with one above and below the flexure lug that control the vertical up- anddown pitch. The remaining two are located at two sides of the center lower limiter and are designed to control the roll of the flexure. These t ravel limiters are fabricated out of Nitronic 60 for material compliancy and high surface wear capability and are designed in pin- like shapes with spherical tips, which were also coated with dry -film lubricant. A back fin extrusion on the flexure end lug common to the chassis bracket was added as contact regi on for the travel limiters. See Fig . 7 for configuration of its design.

Document

AMS_2020.pdf

435 It was a nother lesson learned about the contribution of torsion spring preload to the flexure ’s pitch and roll angles . The flexure to chassis bracket contact from flexure roll was found after DDR with the chassis bracket already fabricated. Therefore, it was a considerable challenge to accommodate the additional roll limiters at this stage of hardware development . When having larger springs attach to hardware with degrees of freedom, the influence on the hardware’s orientation from spring preload should be accounted for. Magnetic Hardstop Similar to how the kickoff spring added additional fail- safe margin and a fault case to the deployment, the magnetic latching hardstop had similar additional benefits for the final deployed position . Due to the larger flexure deployable, there was close clearance concern to the surrounding hardware after deployment , namely the aft bridle tower and its surrounding harnessing on the chassis top deck. A final deployment angle of 40° was chosen to split the difference between the minimum rotation needed by the f lexure to clear the range of motion of the d ifferential and aft bridle tower. From initial testing, there was concern that with the initial hardstop design induced a substantial amount of flexure rebound due to its stiff nature stemming from a C -channel profil e design. After initial contact with the hardstop, the f lexure would then require three re- contacts with the hardstop to come to a complete rest. Around the DDR, the hardstop went through two major redesigns. First, the structural shape was made to be a st andard aluminum sheet metal profile. The final design was the same sheet profile with a bore feature added to house the magnet and its casing (Fig. 9). This diving -board like design added significant compliancy such that the hardstop would dampen out the i nitial contact of the f lexure and reduced the amount of flexure rebound after initial contact . However, during developmental tests, this elastically compliant hardstop rotated nearly 8° over the planned 40° rotational travel for the final deployed position. This exceedance approached the close clearance to the aft bridle tower but did not violate the clearance margin. It was deemed acceptable after being closely monitored at flight level functional testing. Figure 9. Development of the hardstop: (a) initial stiffer C -channel profile, (b) DDR sheet metal design, (c) thinner profile with a feature to house the magnet in the final magnetic hardstop. As a second function, the hardstop is to restrain the deployed flexure once it’s deployed such that the rover differential is completely free to rotate. This hardstop design utilized the flexure’s Maraging 300 steel magnetic properties to make a magnetic latch. The magnetic force of this latch would be added to the residual capability of the t orsion springs. The magnet to be used is of the same material as other magnets on Mars 2020, Grade N42 Neodymium. There was concern about the likelihood of magnet fracture due to its brittle nature directly from impact of the f lexure to hardstop contact . To address this, a protective stainless -steel housing was incorporated. For securement, a ring magnet design was chosen to allow a fastener to run through the magnet and housing and the magnet was to be potted inside the assembled housing (Fig. 10) . Several lessons were learned in the prototyping of the magnet housing. The magnet housing is two separate pieces that when mated together there is a void that would fit the ring magnet. To have the

Document

AMS_2020.pdf

436 magnetic latch work, a magnetic flux circuit with the magnet housing and flexure needs to be closed. Most flight acceptable CRES are austenitic , with the exception of 400 series CRES , which is a ferritic CRES . It was learned that the part of the housing in contact with the magnet should also be the part that is closest to the flexure when the flexure is in contact to the hardstop. This piece of the housing should also be a ferritic steel to complete the magnetic flux circuit from the magnet , through the ferritic housing component, through the f lexure , and return to the magnet (Fig. 10) . This is what produced the latching force. The top of the magnet housing should be an austenitic CRES. Otherwise, the magnetic flux would short through the top part of the magnet housing and not through the flexure. The final design called out the back part of the magnet housing to be 410 CRES and the top part to be 303 CRES. This allowed the magnetic latching force to flow from the magnet through the flexure. The delivered magnetic holding force for the flight unit was measured to be an average of 24.4 N. Figure 10. Cross section of the magnet in its CRES housing with the flexure against the hardstop. Also showing the magnetic flux circuit, which produces the holding force of the magnetic latch. There was a lesson learned that while magnetic latches could be used, some permanent magnet materials cannot tolerate higher temperatures. To comply with mission planetary protection requirements, this neodymium magnet would have seen temperatures during vacuum bakeout such that the magnet could have lost some of i ts residual magnetic strength. As a result, the magnet and its housing were excluded from vacuum bakeout as well as parts of other non -functional tests that went over + 80°C. Planetary protection requirements were satisfied with other methods rather than a vacuum bakeout. In future magnetic latches, it is recommended that the magnet can withstand higher temperatures without any risk of losing magnetic strength over the whole non- operational temperature range. CDPR Design Validation and Verification (V&V) Testing Since t he Mars 2020 CDPR was non- MSL heritage , this new assembly was subjected to a full design V&V campaign that involved a series of functional and structural testing at various lev el of build integration. As of January 2020, all testing was complete d. Rather than presenting the entirety of this lengthy test campaign, two (2) significant tests and some key lessons learned are highlighted. For a complete summary of the test programs in order of execution (including the prototype testing during the engineering development phase) , see Table 1 and Table 2 at the end of this section . Engineering Model (EM) “Dirty” Testing When the CDPR mechanism is fully actuating, in the power decent stage of EDL, there is a lot of Martian regolith that could get kicked up and lodged into the numerous control led gaps within the mechanism. This raised a concern that the deployment of the flexure could be hindered by this “dirty” condition. To address

Document

AMS_2020.pdf

437 this concern, “d irty” testing was performed as a dynamic functional test to ensure its full deployment capability in such an environment. The “dirty” environment was simulated via a media mixture designed to simulate Martian regolith being blown into the CDPR. Using nitrogen gas (GN 2), 700 ml of this media was distributed throughout the entire mechanism and around the mechanism revolute joints. After 30 seconds of circulating the media and thoroughly coating the mechanism, a cord connected to a manual pin (to replace the pin- puller) was pulled to deploy the mechanism. This functional test completed successfully by demonstrating full deployment capability under such “dirty” condition with no signs of hindering. This testing was performed prior to the development of the kickoff spring. This showing that only the redundant torsion springs were sufficient to deploy the flexure and the added kickoff spring only added margin on top of this. To further demonstrate the robustness of the mechanism, a worst -case scenario was also tested where the redundancy of the torsion springs was eliminated and flexure deployment under the “dirty” condition was only driven by the lower and smaller torsion spring. With t hat, the CDPR st ill deployed its flexure successfully with no signs of any hindering. See Fig. 11 for photos of the test configuration. Figure 11. Showing the “dirty” test configuration: (a) showing the before the mechanism was coated with the Mars regolith simulate media, (b) after media circulation and deployment. This test was for the fault case with only the lower torsion spring having successfully deployed the flexure. Engineering Model Loaded Deployment Testing From the rover descent simulation in ADAMS, it was determined that the CDPR would sustain approximately 5.2 kN of residual load at the time just prior to pin-puller firing (Fig. 3). Such amount of residual load would result in approximately 5 joules of potential energy being stored within the fully restrained CDPR. At DDR, this led to a concern that once the pin- puller fires and the flexure becomes unpinned, the then “open” system of CDPR could have an unpredictable and undesirable dynamic response from the sudden release of the strain energy stored in the f lexure. To address this, another dynamic functional test was conducted where the CDPR would be deployed while under 1.2x peak load (6.2 k N through the mechanism) . This “loaded deployment” testing employed a flight -like configuration using a pneumatically driven pin- puller. All springs, magnetic hardstop, and travel limit ers were present in this test and in a flight -like configuration . See Fig. 12 for loaded deployment test configuration. Using a crane and load cell , a lateral load was applied to one side of a Ground Support Equipment ( GSE ) revolute joint to mock the d ifferential CDP, this in turn induced an equivalent load through the CDPR in the opposite direction. A total of five tests were conducted with this setup . Since polarity of such residual load in flight is not 100% certain, t he mechanism would be deployed under a tensile load and then a final

Document

AMS_2020.pdf

438 compression load through the CDPR for full coverage. From the tensile deployments, the mechanism performed nominally. From high- speed video footage, some elastic spring back of the f lexure along its length from the release of tensile strain energy was observed. However, the flexure continued onto its deployment as driven by all the springs. This form of dynamic response was expected and deemed acceptable. During the deployments in compression, however, functional failures were encountered. The following section dives into the test failure o bservation, the respective actions carried out post -failure , as well as lessons learned. Figure 12. Loaded deployment tensile test configuration. The vertical load from the crane is reacted with a linear tensile load in the flexure. The pneumatic pin- puller fires when the crane is at the 1.2x peak load. During this tensile portion of the loaded d eployment testing, it was observed the GSE rotated very quickly about the mock CDP revolute joint due to the strain energy release in the test setup. This configuration resulted in the GSE rotating out of the way of the f lexure deploy path; resulting in no interference in deployment. However, when flexure preload was reversed to compression, such GSE rotation after releasing the pin -puller was also rev ersed. As a result, this reversal caused the GSE to rotate towards the flexure path of deployment and subsequently jammed the flexure; leaving approximately half of the applied preload stored within the system. Once the load was relaxed to nearly zero, the flexure was able to fully deploy. Upon inspection of the hardware post-test, no damage was observed other than minor paint marring on the flexure. To fully ensure this anomaly was truly realistic and repeatable, this compression test was repeated for the final test and same jamming results occurred. The cause of the mechanism jamming was determined to be a function of the test GSE . The GSE rotated two orders of magnitude faster than the fastest predicted flight CDP rotation rate from ADAMS. It is because of this non flight -like, extremely high rotational rate of the CDP , subsequent jamming occurred. More importantly, this jamming event showed that there was a race condition designed into the hardware. This race condition was new and solely problematic for the Mars 2020 CDPR. The overall height of the flexure linkage was big enough to completely block a moving HSA bracket clevis if it does rotate towards the flexure. For MSL, the linkage was only as thick as the spherical bearings and therefore; not big enough to be in any vicinity of the rotational path of the HSA bracket if the linkage was stagnate. To alleviate the concern of this race condition, a simulation of the fault case CDPR deployment against the fastest ADAMS predicted rotating differential was visited. Using as -tested data from CDPR deployments, an analytical model was created that could output the position of the flexure at any point during its deployment. For best accuracy, all environmental conditions such as temperature, orientation, and level of “dirtiness” were considered and properly verified from previous developmental tests. From th is verified analytical model, the outputted flexure positions were compared against the worst -case differential position

Document

AMS_2020.pdf

439 from ADAMS models for the deployment. Results indicated that flexure contact would only occur if the worst -case initial differential of fset and rotational rate are increased by a factor of 2.6 while the CDPR is in this fault case configuration. This was determined to be sufficient margin against extreme environmental cases that the CDPR would win the race condition. Table 1 – EM prototype and developmental testing matrix in order of execution Test Type Test Objective(s) Outcomes EM Flexure Characterization Characterize stiffness at operational/non- operational qualification temperatures and tensile residual strength of the first flexure prototype Successful, no anomalies EM Thermal Functional Testing Demonstrate full deployment mechanism capability at operational/non- operational qualification temperatures in both nominal and fault configuration Successful, no anomalies EM Dirty Testing Demonstrate full deployment mechanism capability while in a dirty environment from regolith kicked up during powered descent of EDL Successful, no anomalies Flight -like Flexure Characterization and Buckling Characterize stiffness and buckling capability up to 1.20x and 1.40x FLL, respectively, at room temperature and respective operational temperatures. Successful, no anomalies Loaded Deployment Testing Demonstrate full deployment mechanism capability under 1.20x FLL Anomaly: jamming observed during compressively loaded deployment, identified race condition in CDPR deployment. Magnet Thermal Testing Characterize thermal effects on magnet installed in the hard stop Successful, confirmed thermal concerns for neodymium Table 2 – DTM and FM V&V test matrix in order of execution Test Type Test Objective(s) Outcomes DTM and FM Flexure Characterization Characterize spring rate of DTM and FM Flexure to 1.2x and 1.0x FLL, respectively at operational/non- operational temperatures . Successful, no anomalies DTM and FM Vacuum Bakeout Thermally eliminate contamination and outgas any volatiles prior to launch Successful, no anomalies DTM CDPR Random Vibration (assembly level) Demonstrate structural capability against protoflight launch environment Successful, no anomalies DTM Therm al Functional Testing Demonstrate full deployment mechanism capability at operational/non- operational qualification temperatures in both nominal and fault configuration Successful, no anomalies FM Thermal Functional Testing Demonstrate full deployment me chanism capability at operational/non- operational hot/cold flight acceptance temperatures Successful, no anomalies FM CDPR Random Vibration (Power Decent Vehicle Level) Demonstrate structural capability against protoflight launch environment Successful, no anomalies DTM Mobility Deploy Demonstrate full dampening of the resultant load during rover asymmetric mobility deployment and full CDPR deployable functionality in earth gravity Successful, no anomalies DTM Rover Chassis and Differential Qualification Static Test Structurally quali fy the entirety of CDPR by demonstrating strength capability up required 1.20x FLL Successful, no anomalies DTM Rover Chassis and Differential Qualification Static Test – CDPR Deployment Demonstrate full deployment mechanism capability under required 1. 20x FLL Successful, no anomalies

Document

AMS_2020.pdf

440 As a worthy note, through this race condition analysis of CDPR deployment fault case, the kickoff spring made the deployment time approximately 33% faster than if there was only the lower torsion spring available to drive the deployment. Such a finding reaffirmed that incorporating an initial motion spring (the kickoff leaf spring) is a great sound design principle. This loaded deployment test was repeated in January 2020 as the last part of the DTM rover chassis and differential design qualification static testing campaign. The CDPR was loaded to 1.2x the limit loads in tension. The flexure was deployed in this configuration using a flight -like pin -puller. No jamming occurred and the same expected dynamics response of the flexure from the strain energy release was observed, resulting in a successful test. Summary For the Mars 2020 mission, a new CDPR achieved similar functionality as the MSL design, while using only half the mass for the structural mechanism by using a new mechanism architecture. The MSL crank -slider mechanism, which was sensitive to misalignment and structural concerns , was replaced with a two -force member spring -like flexure that doubled as providing compliance for the system and as the deployable. This flexure proved to have verified robust design. It was repeatable at temperature and did not require any conditioning. However, this larger deployable required the torsion springs and other hardware of the mechanism to deploy the flexure to be increased in size and strength in order to actuate this larger deployable for the Mars 2020 CDPR . Having the flexure double as the spring dampener and the deployable had the benefit of eliminating the required alignment precision from the MSL CDPR, but also introduced new close clearances and race conditions. The biggest lesson learned was to not have race condition pot ential in a deployable mechanism. It was a large amount of engineering effort to properly show that the mechanism would win this race condition. The analysis on the race condition reaffirmed an important design principle, to always have a simple initial motion spring integrated into a deployable mechanism. Future mechanism could utilize the design of the CDPR f lexure to create a tunable linear spring while ensuring the design does not have the same race condition concerns that were learned from. Acknowled gements This research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with the National Aeronautics and Space Administration. © 2020 . California Institute of Technology. Government sponsorship acknowledged. References 1. Jordan, E. 2012, Mars Science Laboratory Differential Restraint: The Devil is in the Details, Proceedings of the 41st Aerospace Mechanisms Symposium, Jet Propulsion Laboratory, May 1618, 2012

Document

AMS_2020.pdf

441 Astrobee Free- Flyer Nozzle Mechanism Earl Daley* Abstract This paper describes the development and design of the Astrobee Free- Flyer propulsion nozzle assemblies . As will be illustrated in this paper, the Free- Flyer nozzles are thrust control devices used to propel the Astrobee Free-Flyer robot autonomously inside the International Space Station (ISS). The development process, design evolution, and prototyping methods , are described . Key design features are discussed in greater detail to highlight how a seemingly simple design can present surprisingly large challenges . Several lessons learned are given . Introduction The Astrobee project provides free- flying robots (referred to as “ Bees ”), charging station, software, and ground control operations to the ISS for crew support and guest science research. The Free-Flyers are 31.8- cm (12.5- inch) cubes , battery -powered, air-propelled free- flying robot s designed for use within the ISS (Fig 1). Two Free-Flyers (Bees Bumble and Honey) were launched in 2019 along with the charging station (called the “Dock”) and are installed in the Japanese Experiment Module Kibo. A third Bee, Queen, is also on station in storage. Working autonomously or via remote control by astronauts, flight controllers , or researchers on the ground, the robots are designed to complete tasks such as taking inventory, documenting onboard experiments with their built -in cameras or working together to move cargo throughout the station. In addition, the system serves as a research platform that can be outfitted and programmed to carry out experiments in microgravity [1]. Currently, Bumble is going through its commissioning phase and has successfully completed mapping the Japanese Experiment Module in preparation for autonomous operation. Figure 1. Astrobee Free Flyer s and Dock in the ISS * NASA Ames Research Center, Mountain View, CA Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

442 Free -Flyer Propulsion System The nozzles are part of the Free- Flyer propulsion system which is comprised of two modules located on opposite sides of the Free -Flyer (Fig. 2). Air is drawn into the propulsion module by a centralized centrifugal fan which pressurizes a plenum (~1. 5 in-H2O (0.37 kPa) ). Thrust is produced when the pressurized air is exhausted through one of its six nozzles that are located around the plenum. For a constant impeller speed, the thrust from each of the twelve nozzles has fixed direction with magnitude controlled by adjusting the nozzle open area. Figure 2. Free-Flyer Propulsion Module The key propulsion requirements are: 1. provide holonomic control in 6- Degrees -Of-Freedom, i.e., the ability to produce instantaneous thrust in any direction and torque about any axis, 2. produce 0.6- N max thrust on at least one motion axis, and 3. keep noise below 65 dBA at max thrust. The two propulsion modules can be replaced by astronauts and are designed to be interchangeable. Their impeller s rotate at the same speed and, because they are on opposite sides, their gyroscopic moment and drag torques are cancelled. The impeller speed is adjustable to trade thrust performance vs. reduced power and noise. The propulsion layout (dual impellers feeding pressurized air to multiple thrust nozzles ) is conceptually similar to what was employed by the NASA Ames Personal Satellite Assistant project in 2003. Novel features of the Astrobee Free- Flyer propulsion system include low acoustic noise with high thrust , asymmetric nozzle layout which reduc es the total nozzle count while maintaining 6- DOF maneuverability, modular design, and variable thrust control.

Document

AMS_2020.pdf

443 Propulsion Development Process During the early phases of the Astrobee project, several propulsion options were considered. The four leading options were: 1. Onboard compressor : Essentially replicate the Ames SPHERES cold -gas propulsion system ( ref. [2]) except replace the liquid CO 2 tanks with compressed air tanks filled by an onboard compressor. 2. Distributed fans : Use a system of six axial fans with reversible fan rotation, like a multi -roto drone adapted for zero- gee, 3. Similar to option 2, but with Variable- Pitch Propellers (VPP ) to increase fan thrust responsiveness without changing the rotor RPM, and 4. Centralized fans : Use two impellers to drive airflow to many independently controllable nozzles. Option 3, multi -axial VPP fans, was initially chosen for the Free -Flyer propulsi on system. Testing of Prototype 2 is shown in Figure 3. Figure 3. Astrobee Prototype 2 Multi -VPP fan Testing After encountering difficulties with packaging, the VPP system (including limited future payload carrying capabilities) and growing concerns over fan performance and reliability, a new propulsion trade study was started. This time, the ducted nozzles and reaction wheels of the original Personal Satelli te Assistant centralized fan system were redesigned to a single pressurized plenum which greatly simplified packaging. A propulsion system proof -of-concept was quickly made using 3D printed and off -the-shelf parts (Fig. 4) . The system immediately proved capable and was chosen to be developed for flight. Figure 4. Centralized fan Proof -of-Concept Testing on Force Plate

Document

AMS_2020.pdf

444 The basic elements of the centralized fan design are the centrifugal fan, plenum, and nozzles. The propulsion system development was iterative because no single element was independent of , or linearly related to, the other elements and they all needed to work together to meet the final requirements. This meant that iteration occurred at different levels: Component Level, Module Level, and Free- Flyer Level. The challenge was to obtain the data needed, then iterate and then test at the next higher level, then iterate again, and so on. This required significant testing, modeling, and analyses to narrow down the design space. U nfortunately, this started late in the project due to the earlier propulsion system being chosen . Some of the more challenging propulsion system requirements are listed in Table 1. Table 1. Challenging Propulsi on System Requirements Requirement Value Thrust Minimum of 0.6 N in X -direction Acoustics Maximize NC -40 operation (lower noise allows longer operating time) Control Holonomic control Power (electrical) consumption Operate flying ~ 2 hours without recharge Size and Mass Physical envelope and mass limits ISS Requirements Safety, materials, flammability, loads Other project requirements - Replaceable components - Industrial Design - Expanded research options: LED signals, swappable skin Component Level Development Examples of component level tests are shown in Figure 5. Nozzle aerodynamic performance characteristics (along with the impeller) were determined through test using the ANSI Fan Test Rig within the NASA Ames Fluid Mechanics Lab [McLachlan, B., Private Comm unication, 2019]. The test rig, representative test articles and nozzle performance charts based on analysis and test are shown. In each chart, there are multiple curves representing different fan sizes. The left chart shows combinations of fan rotation rate (N) and nozzle open are a (A) that achieve the required maximum thrust (0.3 N for the X -nozzle). The right chart shows the estimated Sound Pressure Level (SPL) for each combination. Figure 5. Examples of Component Level Tests and Analysis Module Level Development Example module level test results and analyses are shown in Figure 6. The photo shows sound and pressure measurements during proof -of-concept testing. Several CFD analyses were completed to study grill effects and parameters such as spacing and depth dimensions. T he chart provides measured SPL vs. Force (Thrust) for two different nozzle areas and four impeller RPM’s. The target SPL and thrust are

Document

AMS_2020.pdf

445 highlighted to indicate operational regions (note, acceptable SPL levels for one propul sion module were estimated to be - 3 dB of two modules in operation). Figure 6. Examples of Propulsion Level Test Results Free-Flyer Level Development Examples of Free-Flyer level test are shown in Figure 7. The granite table test allowed near friction -free 3-DOF testing (2 translational axis and 1 rotation) and was the only “flying” test prior to commissioning on the ISS. As shown, the P4 prototype is mounted on top of a goniometer to test off -axis thrust components. Along with EMI tests, early acoustic measurements were made in the Ames EMI test chamber. Official EMI and Acoustic measurements were made at Johnson Space Center . Nozzle Description Thrust is controlled by varying th e nozzle open area for a constant impeller speed. The nozzle area and impeller are sized to provide the required thrust performance (Maximum 0.3 N for the x -nozzle and ≈ 11m/s exit velocity) . The forward and aft (X -direction) nozzles are physically larger to meet operational thrust requirements. A propulsion module cross -section illustrating airflow and identifying a nozzle is shown (Fig. 8).

Document

AMS_2020.pdf

446 Figure 7. Example of Free-Flyer Level Tests Figure 8. Propulsion Module X -Section

Document

AMS_2020.pdf

447 Nozzle open area is adjusted by two gear synchronized flappers that are driven by an off -the-shelf hobby style RC servo (Fig. 9). The two flappers provid e the benefit of dimensional compactness relative to a single flapper design. The Nozzle grill and divider (made of one- piece 3D printed Ultem 9085) provides flow straightening. Nozzle open area is determined by the flapper angle as shown in Figure 10 (flapper rotation from closed to full open is 64 degrees) . Figure 9. Free-Flyer Nozzle Figure 10. Nozzle Open Area Figure 11 shows a few of the nozzle design iterations with left to right going from the earliest prototype to final flight design respectively . The initial proof of concept nozzle incorporated a single flapper which provided the desired thrust performance but was physically too large in the flow direction to fit the available space. To address that issue, a geared dual flapper design was developed to minimize thickness while providing an acceptable level of flow efficiency . To reduce development time, the first two nozzles were made from 3D printed ABS material. The middle nozzle was an attempt to make a 3D printed “Flight” noz zle (shown is an X-Nozzle), which proved problematic (see Lessons Learned).

Document

AMS_2020.pdf

448 Figure 11. Nozzle Iterations Nozzle Development Issues Several nozzle development issues are expanded on. Nozzle Issue: Flapper Slipping Once calibrated, th e flapper drive elements (consisting of clamps, gears and shafts) must not slip over the life of Astrobee. However, slipping was observed d uring early prototype test s. The required flapper shaft torque is 10 in- lb (1.1 N -m) (which includes a 5X slip factor ). Slipping can occur at two joints (Fig. 12): 1. the flapper to the shaft, and 2. the press fit gear to the shaft. In torque tests, the integrated flapper clamps worked well and were not considered the problem. T he press -fit gear did slip a t low torques . Figure 12. Flapper slip locations

Document

AMS_2020.pdf

449 Option s to increase the torque limits of the titanium shaft to bronze gear included: 1. Increase press -fit interference The nominal fit is ANSI FN2 providing 0001- in (0.025- mm) interference. Torque values as low as 3.68 in- lb (0.416 N-m) were measured in tests . The i nterference was increased to the point that bronze shavings were observed but with little additional torque benefit. 2. Add adhesives Loctite 7649 primer with Loctite 609 retaining compound were tried. Even though the measured torque values were higher than those without adhesive, the decision was made to include option 3 for consistently higher results. 3. Knurl the shaft Knurls are commonly used for small plastic gears on metal shafts ( The cold flow of plastic material into various topographical feature can result in a more constant or increase in torsional strength over time [3]). Also, referencing a common home remedy for slipping where a center punch is used to provide mechanical interlock, a straight knurl on the shaft was considered as an option. The bronze gear would not plastically flow into the knurl, but it could essentially be broached while press -fit to provide an additional mechanical interlock . Several different knurls were tested along with and without Loctite adhesives. The measured torque values were all above 12 in- lb (1.4 N -m) (highest being 21 in- lb (2.4 N -m)). Because of the consistent results, the knurl and Loctite adhesives are used on the flight nozzles. Figure 13 shows the two shafts and Boston gears prior to assembly. The gear in the right image was first press -fit onto the flapper shaft and then removed to show how the bronze gear is cold worked by the shaft knurl. Figure 13. Knurled shafts for improved torque Nozzle Issue: Acoustic Noise and Vibrations Initial nozzle prototypes generated noise that was noticeably loud and annoying. The Free-Flyer’s operating time on the ISS is limited by the maximum noise produced, so it was critical to minimize all acoustic noise sources. The following tasks were completed to reduce nozzle generated noise : 1. Select the best servo The servo was quickly discovered to be the primary noise source. The noise occurred as the servo feedback continuously adjusted to hold a position, with the noise being louder under load. It was observed the noise was different for different servos. The following test was used to select the best servo: With the servo stalled (by commanding it past its limits) measure a ) housing temperature, b) current draw , c) time before failure, and d) a qualitative noise judgement . Seven different servos

Document

AMS_2020.pdf

450 were considered with the MKS DS95 being the clear winner. The MKS DS95 had the lowest noise and was still functional after being stalled for 15 minutes. 2. Remove nozzle backlash springs The initial nozzle prototypes used backlash springs to apply a small closing torque t o eliminat e gear backlash (Fig. 14). Unfortunately , this had an unintended consequence in that the servo overcom es the backlash spring torque produced additional noise. See lessons learned for comments on removing the backlash springs. Figure 14. Backlash Springs 3. Isolate and dampen nozzle vibrations Qualitatively , the servo noise was amplified when mounted to a propulsion module. Also, t he small vibrations could excite surrounding structure and be a motion jitter source for the Free- Flyer. Vibration measurement s of the propulsion system plenum structure, nozzle frame, and servo body were made for an operating nozzle (Fig. 15). Test s with the nozzle hard mounted and isolated, along with the servo mounted with aluminum and nylon standoffs were made. Vibration levels were lowest for the isolated nozzle using aluminum standoffs. The final isolated design is shown in Figure 16. Figur e 15. Nozzle Vibration Test

Document

AMS_2020.pdf

451 Figure 16. Nozzle Isolation Design Nozzle Issue: Bearing Install ation Aligning and pressing the 6- mm and 12- mm O.D. nozzle bearings by hand (and with other simple tools) was time consuming and often resulted in damaged bearings. Including all the flight and ground unit nozzles , over 800 bearings are installed. To improve the bearing installation process , two custom bearing tools were designed and fabricated that proved reliable and capable of rework when necessary. The installation steps are given in Table 2 and illustrated in Figure 17. Table 2. Bearing Installation Steps Step Description 1 Locate Bearing and Nozzle Frame The frame and bearing are radially located by the guide pin and sliding locator (the locator slides on the guide pin and sits on top of the spring) . 2 Press Bearing The guide pin centers the Ram such that only the bearing outer race is contacted as shown in the closeup. The bearing and frame are now conc entrically located and can be pressed together using an arbor press and overcoming the soft spring. The bearing axial position is determined when the Ram bottoms against the frame. 3 Remove Remove the Ram and nozzle frame. Inspect bearing movement. Re-work was necessary when the frame was not held down to the bearing tool allowing a slight angular press . When this occurred, the bearing was pressed through the frame and re- pressed. A better method to securely clamp the frame to the tool would be an improvement.

Document

AMS_2020.pdf

452 Figure 17. Bearing Installation Tool Lessons Learned Several lessons learned are expanded on. Lessons Learned : Limitation of 3D printed plastic parts Significant use was made of 3D printed plastic parts during nozzle development and Astrobee development in general . 3D printed materials make up 32% of a flight Free- Flyer dry mass (the materials include: Ultem 9085, Windform XT 2.0, SOMOS Watershed XC 11122). 3D printed test articles and early prototypes reduced development time and hel ped catch errors in the design. Partly due to the success the team had using 3D printed materials and their lightweight properties, the initial flight nozzles were designed to incorporate 3D printed flappers and nozzle frames (Fig. 11). However, when significant issues with outside vendor part quality and assembly reliability issues were discovered, those parts were changed to a more traditional machined aluminum. The inlet and grill dividers are the only flight nozzle parts made from 3D printed plastic . The Astrobee hardware is classified as non- critical. That meant the 3D printed material strength requirements are driven by the project ’s goal to have a reliable and long- lasting robot , and not th e more rigorous fracture critical material requirements. There are no released NASA standards for materials and parts made of 3D printed plastic (A committee is working on NASA -STD-6030 for Additive Manufacturing, which included 3D printed plastic, but it is not yet released) . A significant effort was required to develop 3D printed plastic standard processes, analysis assumptions, and best practices to assure a level of strength and quality. And in some cases, alternatives to 3D printed plastic shou ld have been explored. A good example is the Free -Flyer painted bezel shown in Figure 1 and Figure 18. A simple design change would have allow ed the part to be easily machined thereby eliminating the additional labor cost necessary to fill and smooth the Ultem 9085 raw surface in preparation for painting.

Document

AMS_2020.pdf

453 Figure 18. Free-Flyer 3D printed plastic painting Lessons Learned: Backlash Testing Gear backlash affects nozzle open area and associated thrust for a given command input . It’s worse for small nozzle openings since the backlash error is proportionally greater for a smaller open area. It wasn’t clear how accurate the open area needed to be during the early development phases . The aerodynamic forces acting on the flappers should pre- loading the gear train and reduce backlash. However, flapper binding was observed on some early nozzle prototypes (which could act in the opposite direction of the aerodynamic forces) and t he decision was made to add a nozzle backlash spring to eliminate all source s of error. A custom 0.012- in (0.30- mm) diameter 302/304 CRES torsion spring with 0.04 in- lb (4.5 mN -m) of torque was added as shown in Figure 14 . The spring torque is applied to close the flappers and help minimize leakage when fully closed (minimizing leaks was another early concern) . The torque direction would minimize the required servo torque since the plenum and airflow pressure s are trying to open the flappers . Two issues developed that focused attention on the backlash springs. The first was the nozzle servo noise as discussed earlier. The backlash spring made the servo noise much worse. The second was the discovery of broken backlash springs on the Prototype 4 (P4) Granite Table Testing shown in Figure 7 . Smal l 0.012- in (0.30- mm) diameter spring pieces floating in the ISS cabin would be an obvious safety concern. The decision was made to try operating P4 without the nozzle springs. The P4 testing had been successful and going on long enough that any adverse changes to the Free- Flyer could be measured. The tests showed no change without the springs and the servo noise was noticeably lower . The backlash spring was removed from the design (however, the flapper still has a notch for the spring due to fabrication schedule constraints ). The backlash springs were added because of uncertainty in the required nozzle opening accuracy and nozzle function. As it turned out, the springs were added to solve a theoretical concern that wound up causing a real pro blem. It was fortunate that the design was not impacted when they were removed.

Document

AMS_2020.pdf

454 Conclu ding Remarks This paper describes how complex the nozzle design was for something that overall is rather simple. It’s typical that designs quickly become boxed in by competing requirements and practicality. But t he added dependency of full system test to feedback the next design iteration made this more difficult and schedule critical. The original project plan included four (4) granite table “flying” prototypes (Free -Flyer Level prototypes) . The prototype (iterative) philosophy was a good choice for this development. However, because the centralized fan wasn’t the original propulsion system choice, it wasn’t until Pr ototype 4 (P4) that it could be fully tested and provide valuable data for the next iteration. The centralized fan propulsion system has proven itself capable and a high-performance option for intravehicular robots. Future centralized fan propulsion designs can be scaled and optimized. The Astrobee Free-Flyers serve both as a feasibility demonstration and reference design for future mission- critical intravehicular robot systems. References [1] website https://www.nasa.gov/astrobee [2] website https://www.nasa.gov/spheres [3] Malloy, Robert A., Plastic Part Design for Injection Molding, Hanser/Gardner Publications, C incinnati , OH (1994).

Document

AMS_2020.pdf

455 Major Design Choices and Challenges that Enabled the Success of the Ejectable Data Recorder System Jeff Hagen* , Michael Burlone* and Kristina Rojdev* Abstract The Ejectable Data Recorder (EDR) subsystem was added to the Crew Module (CM) of the Ascent Abort 2 (AA- 2) test flight due to a risk that the communications architecture would be insufficient to downlink all the data to the ground during the test flight. S ince the EDR subsystem was a secondary system for data collection, AA -2 management enabled the team to take a different approach to hardware development that was more agile- like. This paper discusses key design decisions , technical challenges, and lessons learned that enabled the success of the EDR system during its flight on July 2, 2019 . Background The successful Ascent Abort 2 (AA -2) test flight on July 2, 2019 demonstrated the Orion Launch Abort System (LAS) performance during the maximum dynamic pres sure phase of the launch ( Figure 1). In order to reduce costs, the test used an aerodynamically and structurally representative boilerplate substitute for the Orion spacecraft Crew Module (CM), which was not equipped with a parachute recovery system. Rather, the CM was discarded several miles off shore at the conclusion of the test. Although data from the developmental flight instrumentation (DFI) system was returned via high data rate telemetry link, the risk of interruption in the telemetry justified the development of a backup data recording system. T he CM splashed down into the ocean after the launch and was not intended to be recovered. Thus , the data recorders were ejected after completion of the LAS test phase of the flight , but prior to water impact of the CM . Then, the data recorders were independently retrieved via water recovery operations . Figure 1. AA-2 launch. The Ejectable Data Recorder (EDR) subsystem had three major functions: record the data, eject the payloads and provide the location of the payloads so that they could be retrieved. These functions were implemented through a dual string redundant system that consisted of a US Air Force AN/ALE -47 flare and chaff dispenser system, a breakout box that contained the computer for the recordi ng capability and routed power and data, and a payload ( Figure 2) with a memory device and a beacon. The details of this design * NASA Johnson Space Center, Houston, TX; [email protected] Proceedings of the 45th Aerospace Mechanisms Symposium, NASA J ohnson Space Center, 2020

Document

AMS_2020.pdf

456 will be presented at the IEEE Aerospace Conference in March 20201. This paper expands upon the previous paper with focus on key design decisions and a non- traditional development process at NASA that enabled success of the EDR subsystem , within the constraints of limited resources and a compressed schedule. Figure 2. Six of the twelve flight payloads near completion. Aggressive cost and schedule limits for the AA -2 program, exacerbated by the late addition of the EDR subsystem, forced development to proceed along an accelerated path with resource limitations that did not allow for a typical process of developing requir ements, defining environments, and analytically validating design solutions. With an incomplete set of requirements, engineering judgment guided the identification of design parameters with the most uncertainty, and/or highest risk of not converging on an acceptable solution. These parameters were pursued with demonstration testing as early as possible in the development process, which provided the necessary evidence regarding those requirements with the highest sensitivity to the design parameters . Require ments definition matured concurrently with testing progress . The design was explicitly driven to maximize potential for large margins in parameters with high uncertainty and low sensitivity. The design also allowed for radical redundancy – twelve data recorders were flown, but only one needed to be recovered to obtain the full flight data set. Unique Challenges The principal challenge was to determine how to eject a recorder from the CM. Detailed knowledge of the local environments at the time of ejection was not available, but the ejection system needed to ensure confidence in successful retention and separation of the data recorder payloads during the anticipated tumbling of the CM during uncontrolled flight. Due to the cost and schedule constraints of the project, making use of an off the shelf solution was preferable to a new development. Early on it was recognized that the US Air Force AN/ALE -47 flare and chaff dispenser system had many features that seemed to be congruent with the EDR objectives. The standard ALE -47 dispenser system required an enhanced capability in two critical aspects. The anticipated flight vibration environment of the CM during the test flight greatly exceeded certified limits of the ALE -47 system. In addition, the ALE -47 syste m was designed for dispensing passive decoy payloads and did not have provisions for separable power and data connections between the payloads and host vehicle. Thus, modifications to the ALE -47 system were required to meet these needs for the AA -2 test flight. Bounded by the standard capabilities of the ejection system, the defined characteristics of the payload data storage and beacon electronics, and the requisite operational environments, the dominant aspect of the payload design was the balance of pac kaging the components into the available mass and volume limits of the standard ALE -47 cartridge, while still maintaining adequate robustness. Not only did multiple electronics need to fit within the volume allotted by the cartridge, but the mass was const rained by the need for sufficient buoyancy to ensure the payloads would float in the ocean with adequate stability for antenna pointing, yet have sufficient structural integrity to survive the water impact at a terminal velocity of approximately 150 mph.

Document

AMS_2020.pdf

457 Ejection System Concept In the ALE -47 system, the payloads were packaged into cartridges that function similar to a mortar tube (Figure 3). Crimping of the open end of the aluminum cartridge case restrains an aluminum closeout plate (cap) that retains the payload in the cartridge until ejection. A specially designed machine performs the crimping operation, which guarantees that the payload has no residual gap and cannot move within the cartridge. An electrically initiated pyrotechnic squib installed in the base of the cartridge provides the propelling gas that forces the payload and closeout plate through the restraining crimps and out of the cartridge. A plastic piston assembly installed in the bottom end of the cartridge ensures that the combustion gasses from the squib do not leak past the payload and negate the propelling force, analogous to the function of the wadding in a shot gun shell. Figure 3. The figure on the left shows the components for assembling the payload into the cartridge and the figur e on the right shows the pass thru plugs installed into the payload interface. The cartridges were packaged into a magazine that constituted the operational load element of the system, and were installed on the vehicle by sliding into the muzzle of the ALE -47 dispenser after installation of the pyrotechnic squibs ( Figure 4). The modularity of the dispenser and magazine system enables installation of as many as 30 cartridges of varying sizes with payload cross sections ranging from 1” x 1” up to 2” x 2.5”. Only six of the largest size cartridges required for the AA -2 EDR payload can fit in each dispenser, which leaves 24 of the electrical squib contacts in the dispenser breech plate unused and dead faced against the cartridge bases. Figure 4. The fi gure on the left shows the payload packaged inside the cartridge and the cartridge being installed in the magazine. The first on the right shows the assembled dispenser next to the breakout box installed in the forward bay area of the CM.

Document

AMS_2020.pdf

458 Two independent ejection systems were mounted in the forward bay parachute compartment of the AA -2 CM, pointing in opposite directions ( Figure 5). Figure 5. EDR ejections. Figure on the left is an artist rendering and the first on the right is an image from the actual f light. Risk Definition Risks were based on perceived probability and consequences, but also ranked according to the expediency or difficulty of accomplishing closure. Tests that could be conducted rapidly and efficiently received schedule preference in order to identify pitfalls as soon as possible. T he principle risks identified were as follows: • Functionality of the COTS ejection system and EDR unique payload • Adequate separation performance of the ejection system in the poorly characterized post -test flight environment • Mechanica l survival of the payload during flight, ejection, and impact environments • Recording of the data stream and storage of the data in the payloads • Separation of the power and data connections of the payload through the ejection system • Packaging of the payload elements • Buoyancy and stability of the payload • Location and recovery of the payloads • Survival of the system in the flight vibration and shock environment • Payload battery performance and survivability • Beacon activation Critical Developmental and Risk Redu ction Testing Project resource limitations drove testing concepts to prioritize low -cost approaches using existing or readily available, and often non- traditional, assets to the fullest possible extent. One of the earliest tests was performed using an air craft and recovery boat to drop and recover inert simulated payload shapes into open water. The simulated payload shapes were constructed from a prototype syntactic foam mixture and were dropped from an altitude well in excess of that necessary for the pay loads to reach terminal velocity. This testing proved that syntactic foam provided sufficient protection to the payload. This test was followed by an ejection from the ALE -47 dispenser using inert simulated payloads . With the data from this initial testing and multiple iterations of design and 3D printing, the team began to close on a payload design that encompassed the volumetric, buoyancy, and structural constraints.

Document

AMS_2020.pdf

459 The next payload prototype included a n early generation, functioning locator beacon and recording computer, contained within a 3D printed plastic payload shell ( Figure 6), as opposed to an alumi num shell used in the initial design. Ultimately, this shell printing process proved vital to the frequent and rapid iteration of the payload design, the extremely low production costs and quick fabrication times . 3D printing also enabled the project to construct the numerous development and replacement payload units needed to work through many test failures , as well as the large number of flight units for redundancy purposes . The ejection performance test was repeated with this enhanced payload, and showed that the ejection system could adequately eject a functionally complete payload wit h sufficient velocity, and that the payload could survive the ejec tion forces. Figure 6. The image on the left is a set of 3D printed payload shells and scaffold. The image on the right is an assembled payload that is prepared for foam pouring. Collaborative use of a NASA- owned Gulfstream III aircraft equipped with an instrument package deployment capability allowed for early demonstration of water impact survivability and recoverability of prototype active payloads. Five payload prototypes of varying configuration were dropped from approximately 5000 f eet (1.5 km) above sea level and approximately 50 miles (80 km) off shore from Galveston, Texas. Beacon signals were received from all units, although two units failed to send proper GPS coordinates . These discrepancies were thought to be due to unsuitable antenna configurations or assembly defects. The other three units were recovered and their simulated data retrieved (Figure 7) . Figure 7. Early drop testing and retrieval operations. Following were data writing tests, initial demonstrations of payload power and data harness separation, vibration testing of the ejection system elements, tests of data writing during vibration, increasingly complex payload recovery operations, thermal exposure followed by ejections (Figure 8), ejection system performance with fully active payloads, payload beacon battery failure margin testing and recovery beacon activation tests.

Document

AMS_2020.pdf

460 Figure 8. Ejection testing of a data recorder mass simulator. Dispenser Vibroacoustic Mitigation The expected vibroacoustic environment during the abort sequence of the AA -2 mission greatly exceeded the rated capabilities of the ALE -47 dispenser (Figure 9). The system specifications of the off -the-shelf dispenser accommodated a lower amplitude vibroacoustic environment, but for a much longer duration than the abort sequence (3 hours for the benign rated environment, as opposed to 5 seconds of the AA -2 abort environment). Figure 9. Dispenser rating compared with the AA -2 abort environment. Several mitigation strategies were necessary to produce a design that could reliably withstand the extreme environments induced by the abort motor . These strategies included using additional structural attachment points to the dispenser and mounting the dispenser on cup- style elastomeric isolators ( Figure 10) . Notably, the dispenser with its custom mounting interface was tested to the extreme abort environments while mounted on hard mounts (i.e. solid aluminum stand- ins for the isolators) and survived. However, with the dispenser in the hard- mounted conf iguration, data transfer to the data recorders loaded inside was unsuccessful.

Document

AMS_2020.pdf

461 The off -the-shelf dispenser provides four attachment points near the muzzle of the dispenser to attach to vehicle structure (Figure 10 ). The center of mass of the dispenser is not in plane with these attach points, resulting in a cantilevered arrangement. To avoid this cantilevered arrangement, the AA -2 custom design made use of the subassembly fasteners near the breech of the dispenser to affix plates and brackets such that the custom dispenser mounting interface was nearly coplanar with its center of mass. Figure 10. The image on the left shows the four attachment points. The image on the right shows the suite of dispenser modifications. These additional brackets were located in such a way that when mounted on isolators, the center of mass of the dispenser was near the centroid of the isolators . This design eliminated rotational coupling (i.e. swaying motion) of the dispenser when exposed to vibrational environments, keeping the resulting motion almost entirely translational . The isolators themselves were selected such that the natural frequency of the isolated mass would be approximately 32 Hz, providing vibroacoustic attenuation for frequenci es above 45 Hz . Figure 11 shows the frequency response of the isolated dispenser based on the theoretical transmissibility of the selected isolators. Figure 11. Theoretical Frequency response of isolated dispenser.

Document

AMS_2020.pdf

462 Even with the isolators, the dispenser experienced a vibroacoustic environment above its rated values for frequencies below 200 Hz . However, with the customized mounting arrangement of the dispenser and brief exposure to the abort loads, the system design passed qualification testing and was ultimately successful in the AA -2 Mission. Lessons Learned During Vibroacoustic Testing Cartridge Crimp Failure An early vibration test had a failure of the crimped cartridge to contain the ejectable data recorders . Ideally, the contents of the cartridge would be preloaded in a manner that resulted in no backlash or motion of the contents of the cartridge. However, an error in the crimping procedure resulted in a free displacement of the recorder between .007 in (0.18 mm) and .044 in (1.1 mm) within the c artridge. Even this small motion was enough to produce significant impact loads on the crimp feature, which was defeated. During the test, the lids came off the cartridges, and due to the asymmetric friction resistance of the piston seal (i.e. higher force needed to push seal in than to pull out), the vibration motion “ratcheted” the piston and data recorder out of the enclosed cartridge. Recalibration of the crimping machine at the cartridge loading vendor’s facility, along with inclusion of a felt spacer and a bead of RTV sealant as used in the standard cartridge configurations, resolved the loose fit issue such that this failure did not reoccur ( Figure 12). Figure 12. Cartridge crimp failure during vibration testing. Dispenser Isolator Failure Another test failure occurred on the dispenser when an erroneously entered test profile was programmed to the shaker table, which resulted in the destruction of the isolators . After the failure occurred, it was determined that the erroneous profile differed from the intended profile as shown in Figure 13. The 32 Hz frequency is notable since that is the system’s natural frequency based on dispenser mass and isolator stiffness.

Document

AMS_2020.pdf

463 Figure 13. Figure on the left shows the erroneously tested vibration profile. The images on the right show the resultant cup isolator following the test. Curiously, the g rms value of the erroneous profile that destroyed the isolators was 17% lower than the g rms value of the intended profile (29 g rms vs. 35 g rms), even though the isolators were proven to survive the intended profile when tested on a later date. In this case, the erroneous profile exceeded the intended profile in the most crucial frequency range, the low frequency domain, which included the system’s natural frequency of 32 Hz (Figure 14). Applying the isolators’ transmissibility to calculate the frequency response of the isolated dispenser creates a clearer picture of why the erroneous profile was an over -test. Figure 14. Dispenser frequency response to er roneously tested vibration profile. As before, the g rms value does not tell the full story . The frequency response of the isolated mass shows only a slight increase in g rms when comparing the erroneous profile to the intended one (12.1 g rms and 10.9 grms, respectively) . The true significance of the exceedance in the low frequency is not apparent until examining the rms displacement, instead of the rms acceleration. Acceleration values resulting from oscillating motion are linearly proportional to displac ement amplitude and proportional to the square of the oscillation frequency . This means, for example, that for a body

Document

AMS_2020.pdf

464 experiencing a given rms acceleration from sinusoidal motion at 400 Hz, the peak -to-peak displacement when experiencing the same rms accel eration at 40 Hz is 100 times greater . Thus, given the two acceleration profiles, with similar g rms values, shown above, it is expected that the profile with a higher acceleration spectral density in the low frequency domain, namely the erroneously tested profile, will have a greater rms displacement. Indeed, when comparing the dispenser’s frequency response displacement from the erroneous profile, with its response displacement from the intended profile, it becomes clear that the erroneous profile was muc h more severe. Figure 15 shows the comparison on a non- logarithmic plot. Figure 15. Dispenser frequency response displacements to erroneously tested vibration profile. The rms displacement of the intended profile is 0.107 cm, while the rms displacement of the erroneous profile is 0.299 cm. The test failure that occurred was a result of excessive displacement of the isolators . It is notable that although the erroneous test profile indicated a 17% more benign acceleration value (g rms), the resulting displacement that the isolators experienced was almost triple the intended value. Data and Charging Link Challenges The standard ALE -47 system dispenses only passive payloads and does not include any provision for data or power interfaces betwe en the host vehicle and the payloads, as required by the AA -2 EDR application. The power connection served only to charge the EDR batteries prior to launch and therefore did not function during the flight environment, but the data connection had to maintai n integrity of the 20 Mbps real time data stream between the local data controller and the USB solid state memory embedded in the EDR throughout the full flight load, vibration and EMI environments. Additionally, these connections had to be separated upon ejection in a manner that preserved the data connection for use after recovery. Each of the two connections required four independent conductive paths. The field installation nature of the ejection system pyrotechnic devices required that the payloads be i nstalled and thus these connections be made for the final time only during the later stages of the launch preparation process, long after the EDR system had been built and installed in the host vehicle. There were two concepts traded to sever the data and power lines between the vehicle and the payload. The first concept employed a blade in the cartridge that would cut the wires as the payload ejected, similar to a guillotine. The other concept routed the power and data lines through connecting plugs between the payloads and dispenser, oriented with their axis of separation aligned to the axis of payload ejection. After

Document

AMS_2020.pdf

465 several different tests and trade discussions, the concept that employed a connecting plug interface was selected due to the ability to eas ily insert and remove the populated magazine from the dispenser multiple times. Ejection of the payload pulled the plugs from their associated jacks. Routing of the cables to the external interfaces took advantage of the excess pyrotechnic squib electrical interface locations in the dispenser breech plate not required for use in the six payload configuration, such that installation of the loaded magazine into the dispenser accomplished engagement of the plug. Because the gas expansion chamber formed by the piston and the base of the cartridge case fully encompassed the end face of the payload perpendicular to the axis of ejection, the umbilical connections had to pass through both sides of the gas expansion chamber without introducing an excessive leak path. Due to the imprecise nature of the EDR construction, the positioning of the connecting jacks within the payload were not tightly controlled. The relative positioning of the payload connectors to the cartridge base, the cartridge base to the dispenser breech plate, and the connector location to the piston seal constituted three independent alignment challenges, each with six degrees of freedom, that had to be accommodated by a separable umbilical connection routing through the cartridge base and dispenser breech. The connection was required to prevent separation or electrical chattering of the plugs under all conditions prior to payload ejection. The design solution employed four conductor, coaxial mini -phono connectors as the separable plugs due to their axisymmetric nature in eliminating the need to accommodate rotational alignment constraints, self - alignment of plug insertion, robust connection strength, compact size with high conductor density per cross section area, and low cost. The precise fit nature of these connectors required compliance for each plug to follow the unique lateral, axial, and angular alignment of its corresponding jack. In order to simplify the combined alignment challenges, each separable plug connection was split into two different plug and jack pairs, with one pair constituting the mating interface for the loaded magazine installation function and the other pair constituting the demating interface for the payload ejection function. Due to the nature of the payload assembly within th e cartridge, it was necessary to have a separable connection at the payload to cartridge interface, and another one in series at the cartridge to dispenser interface. The results was the “pass through plug” ( Figure 16), which consisted of a mini phono plug on the payload mating end linked to a mini phono jack on the dispenser mating end. Figure 16. Pass through plug and floating plug installation.

Document

AMS_2020.pdf

466 The connectors of each pass through plug were embedded in a two- part shell capable of accommodating the compliance necessary to ensure alignment of the assembly , such that the plug was fully seated and the shell adequately aligned, in order to seal in the ejection gas generated by the pyrotechnic charge. This plug and jack pair constituted the demating interface for the ejection function of the connectors. Oversized holes in the piston and piston seal accommodated alignment compliance between the installed pass through plug assembly and piston assembly. A layer of compliant seal material (rubber sheet) , sandwiched between the upper surface of the piston and the lower end of the payload, provided a seal against leakage of the expansion gas without requiring a tight fit between the pass through plugs and associated holes in the piston and piston seals. Phono plug assemblies (the “floating plug”) installed in the dispenser breech plate, in place of the squib electrical interface pins , constituted the mating interface for the magazine installation function of the connectors ( Figure 16). Within the dispenser, a compression spring at each connection provided enough force to guarantee full insertion of the connections once the payloads were fully seated, and allowed for flexure of the connectors to accommodate any misalignment ( Figure 17). Figure 17. Dispenser to payload pass through plug interface. The original intent was to limit the payload installation into the vehicle to a single cycle, but unanticipated challenges with the software of the data recording and payload beacon activation forced an extended test campaign with numerous installation and removal cycles. The environmental test plan of vibration, thermal, and ejection test s also required a larger number of payload installation and removal cycles than originally envisioned. The repeatable and simple nature of this connector concept proved both vital to testing, but also problematic. The inherent nature of the coaxial plugs caused a variety of adverse connector pin shorting conditions during insertion and removal that required development of detailed and rigorous operational procedures to avoid damage to the payloads. Although initial data integrity testing appeared successful at the component level, early development tests of data writing during vibration encountered anomalies with data integrity. Initial diagnostic efforts focused on the apparent association with the vibration environment, but further testing revealed that intermittent data writing failures exist ed when using the fully integrated system configuration in any environment. Si gnal diagnostic testing uncovered that although data had been successfully transmitted and the combined connection had excellent conductivity on all lines, when operating at the high frequency and low signal strength of the USB 2.0 standard, the fully assembled data connection had a net impedance sufficient to cause intermittent failure of the data signal. Upon comparison of the coaxial phono plug connector with traditional USB connectors that have individual, parallel signal paths, it was realized that the closely spaced concentric geometry of the coaxial contacts in the phono plug created a capacitance between adjacent

Document

AMS_2020.pdf

467 connector rings. With the positive and negative high frequency data channels placed on adjacent rings, this capacitance resulted in an impedance at the high frequency of the data signal sufficient enough to block the low voltage signal when combined with other losses in the complete system. By reordering the USB channels through the connector such that the ground channel of the USB DC power bus was placed on the ring between the two data channels, the capacitance effect was disrupted and the impedance problem eliminated. Beacon Activation and Battery Maintenance Challenges Since it was necessary to avoi d premature emission of radio signals and draining of battery capacity, the locat or beacon was not activated until the payload was ejected. D ue to the extended period of dormancy from final battery charge until payload ejection, it was necessary to minimiz e the power draw of the ejection detection system in the payload. The EDR payloads included a custom designed activation circuit for the ejection detection and battery charging functions. The coaxial nature of the phono plugs was susceptible to brief short ing between adjacent conductors during insertion and removal , which in combination with peculiarities in the activation circuit design, led to a couple of challenging problems uncovered during later phases of testing. Early versions of the activation design involved means of detecting severance of the USB data connection circuit via cross connections to the battery charging circuit. T his cross coupling resulted in numerous permanent failures of the payloads during testing, particularly to the COTS flash dr ives used for data recording. Eventually it proved impossible to eliminate the failures with the original concept . A new system of detecting payload ejection was developed. This method sens ed the severance of a short circuit between terminals protruding fr om the side of the payload and shorted through contact with the wall of the alum inum cartridge prior to ejection. This late design change succe eded in eliminating the flash drive failures due to problematic coupling of the charging and data circuits . Due to inadequate voltage specification of certain components in the activation circuit , another consequence of crossing conductors during coaxial plug insertion and removal was a propensity to damage the activation circuit or battery, which accumulated over r epeated cycles. Delayed understanding of the issue precluded redesign for properly sized components. Thus, the vulnerability was mitigated through procedural safeguards and rigorous monitoring. A final redesign was undertaken to completely replace the act ivation circuit with an optically sensed circuit for detecting payload ejection. While this completely resolved all of the remaining risks for operational damage or premature battery drain of the payloads, the very late introduction in the schedule precluded the opportunity for full adequate testing of this design. In the end, a mixture of both redesigned versions of the payloads was flown in order to balance these conflicting risks without disrupting the flight schedule. Conclusions The most challenging development issues were related to the custom designed payload activation circuit and the software for writing data to the flash drives. Ironically, neither of these had been considered noteworthy risk items at the beginning and had assumed to be within t he scope of routine developments. In hindsight, the one noteworthy improvement to the development process would have been earlier acceptance of the need for a change of design concept for these two items and better integration between mechanical and electr ical aspects of the design process in the crucial early phases . In operation, the EDR system performed almost flawlessly. It met all critical schedule milestones up to and including flight , and without disruption to the overall AA -2 development and launch preparation schedule. During the mission, all 12 data recorder payloads ejected exactly as planned and were successfully recovered ahead of the anticipated timeline. Data was successfully retrieved from all units and ultimately proved to be a perfect reco rding of the telemetry stream on 11 of the units, with the only flaw being limited disruption of the data on one of the payloads. The data loss was an intermittent issue with one of the data

Document

AMS_2020.pdf

468 channels of the breakout box noted during testing, but accepted d ue to the impracticality of repair and protection of massive redundancy. Other than the noted issues with early versions of the payload beacon activation circuit and data writing software, the programmatic and operational performance of the system fully validated the selected design choices and development philosophy of the EDR subsystem. References 1. Rojdev, K., Hagen J., Burlone, M., Petri, D., Jackowski, A., et al., “An Ejectable Data Recorder Subsystem for the Ascent Abort -2 Test Flight of the O rion Launch Abort System,” Proceedings of the IEEE Aerospace Conference, Big Sky, March 7- 14, 2020.

Document

AMS_2020.pdf

469 Design and Test of the Orion Crew Module Side Hatch Lance Lininger* and Kyle Gotthelf** Abstract The Orion spacecraft is a critical component of NASA’s Artemis program to establish a permanent human presence on the lunar surface and further enable future crew ed missions to Mars. As a result of traumatic lessons learned from the Apollo and space s huttle programs, NASA has established thorough crew safety requirements that were particularly challenging to implement for the Orion Crew Module side hatch. This paper describes the background and evolution of the side hatch design, the features used to address the crew safety requirements, the testing challenges in preparation for the Artemis 1 mission, and lessons learned. Introduction Due to four decades of hard- learned crew safety lessons (Apollo 1 command module fire, space s huttle Challenger, and space s huttle Columbia), NASA established requirements to make Orion the safest and most reliable human- rated spacecraft ever built . However, these re quirements when coupled with the long Orion mission durations, introduced some significant challenges with the side hatch design (see Figure 1 for the location of the side hatch on the Orion Crew Module) . Figure 1. Orion Crew Module Specific side hatch changes from Apollo/ space s huttle include (also see Table 1): • Introduction of the NASA -STD-5017 mechanisms design standard for crew ed missions (based on AIAA- S-114) • More conservative structural/force/torque margins • Implementation of redundancies (where possible) • Higher factor on mechanism life cycle verifications (4X for crewed missions) • Longer mission duration • Higher cabin pressure • Lower leak rate * Lockheed Martin Space, Littleton, CO; [email protected] ** Honeybee Robotics, Longmont, CO Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 2020

Document

AMS_2020.pdf

470 Table 1. Side Hatch Key Driving Requirements Comparison, Apollo vs. O rion Requirement Apollo Orion Cabin Pressure 44.8 kPa 106.9 kPA Max Leak Rate ~2.3 kg / day (estimated) 0.68 kg / day Crew member size/strength Average M ale range 1% female to 99% Male Seal Redundancy Single seal Redundant Seals Hatch Size 0.64 m2 0.79 m2 Hatch Opening Cycle Life Records uncertain 1180 cycles Mission Duration 14 days > 21 days NASA -STD-5017 (AIAA -S-114 derivative) applicability Didn’t exist Yes As a starting point, t rade studies were conducted to identify the design solutions that were most likely to meet these requirements. Past crew hatch designs were examined from Gemini, Apollo, space s huttle, and ISS. After numerous reviews and discussions, it was determined that a highly modified version of the Apollo hatch was best suited for Orion. Figure s 2 and 3 show the Apollo and Orion side hatch designs with key components. Figure 2. Apollo Side Hatch

Document

AMS_2020.pdf

471 Figure 3. Orion Artemis 1 Side Hatch Orion Side Hatch Design Details The side hatch contains the following key components: • Aluminum structural panel : An initial trade was performed on a flat vs. curved structure (Apollo used a curved structure). Due to the more stringent sealing requirements, it was decided that a f lat hatch structure would be used on Orion (to improve sealing tolerances and reduce cost). It was also theorized at the time this would simplify the latch train kinematics (latches wouldn’t have to transmit loads over a three- dimensional curved surface), although this turned out to be a negative aspect of the design later in development testing. • Latches: Two latch trains ( “A” train and “B” train, 17 latches total) are used to compress the perimeter seals and retain the hatch to the spacecraft structure. Due to increased strength margins and higher pressures required on Orion, the latches were scaled- up significantly over Apollo (see Figure 4). The latches are over -center mechanisms , linked together through a series of linkages and bell cranks . Ideally, as hatch pressure loads increase this should drive the latches f arther over center into the hardstops (see Figure 5). Rigging of the latches is critical. I n a perfectly rigged hatch, all the latches would be overcenter and contacting the hardstops simultaneously . However , even with designed- in adjustable rigging features, it is extremely difficult to achieve the desired hardstop conditions due to the many part tolerances and the interactive nature of the latch linkages.

Document

AMS_2020.pdf

472 Figure 4. Apollo vs. Orion Latch Size Figure 5. Latch Over -Center Action • Deployment hinges : 4-bar linkage design (Figure 6) . The opening motion is optimized to provide the desired final hatch position needed for crew emergency egress, while also providing an appropriate closed position for sealing. Figure 6. Side Hatch Hinge

Document

AMS_2020.pdf

473 • Gearbox: Honeybee Robot ics m ulti-stage gear system used as the interface for both ground and flight crews to actuate the latch trains from inside and outside the Crew Module (Figure 7). A key gearbox feature is the Latch Restraint Arm (LRA), which is a spring- loaded structural element that engages the root of both latch trains when the latches are in the final over -center position ( Figure 8). This acts as a redundant structural load path to prevent inadvertent unlatching. The gearbox include s an LRA switch that arms and disarms the spring mechanism which places the LRA into the latched position (see Figure 9 for the original switch design used on the side hatch development unit). The Arm/D isarm Switch is meant to serve two primary functions. One of the needed functions is to ‘disarm’ the LRA from the interior of the Crew Cabin so the astronauts can open the side hatch via the Interior Latch Actuation Handle (ILAH). Disarming the LRA is done by rotating the switch in the counterclockwise direction. The act of disarming the LRA can also be achieved from the exterior of the Hatch via ground crew actuating the Exterior Latch Actuation Mechanism (ELAM). The other primary function of the switch is to ‘arm’ the LRA by rotating the switch in the clockwise direction. This is meant to be performed prior to closing the hatch. Therefore, the act of arming the mechanism is meant to be an ‘action’ and not a ‘state’. Arming the switch will allow the LRA to ‘engage’ and lock the latch train but this will only happen when the latches are actuated to the fully closed position. When the LRA engages it is meant to ‘kick’ the switch back to its Neutral position. Figure 7. Gearbox Overview

Document

AMS_2020.pdf

474 Figure 8. Gearbox Latch Restraint Arm (as viewed from outside the hatch) Figure 9. Original LRA Switch Design on the Development Hatch

Document

AMS_2020.pdf

475 • Counterbalance assembly: Moog Inc. hydraulic/pneumatic actuator which provides the motive force to open the hatch in both normal and emergency egress conditions once the hatch latch trains are disengaged (Figure 10). Figure 10. Counterbalance Assembly Initial Design and Analysis The Artemis 1 mission (which is uncrewed) did not originally include a full fidelity side hatch but rather a structural simulator for the flight . It was decided late in the Artemis 1 design s chedule to add a full fidelity side hatch to reduce technical risk prior to Artemis 2 (first crewed mission) . This decision accelerated the side hatch schedule by approximately one year which introduced obvious programmatic challenges. Due to these schedule constraints the design team had to be judicious on the level of detail in the Finite Element Model ( FEM ) and kinematic analyses. In order to meet the Critical Design Review and the development testing milestones, t he initial analysis of the hatch design was limited to standard FEM/stress modeling and rigid-body kinematic studies of the latch trains which all showed positive results with significant margins. There was a plan to do a mor e in-depth flex -body analysis to examine the complete interactions of the structure and latch trains, h owever this couldn’t be completed in time. The CDR was held successfully (with many independent NASA and LM reviewers) and so after n umerous meetings wit h management , it was agreed the team should proceed into development testing to validate the design. Development Testing As typical for a mechanical system of this complexity, the team started tests at a subcomponent level. Numerous structural and life tests were conducted on the latch mechanisms, including individual latches and a small connected set of 4 latches. The static load test of a single latch is summarized in Figure 11. This test was intended to be an ultimate load test, and the actual capability was at least 2X the anticipated strength (the test was stopped due to limitations of the test fixture) . This gave the team confidence that the latches were very structurally robust to react the pressure loads and would easily meet the more stringent NASA structural requirements. The test also validated the over -center action of the latch (at an individual latch level).

Document

AMS_2020.pdf

476 Figure 11. Individual Latch Pressure Simulation Static Load Test After various subcomponent tests, a full -size flight -like development hatch was built and a development test program was implemented. The initial plan for development testing included: • Functional testing • Proof -pressure testing (at 1.5 x nominal mission pressure) • Vibration testing (at pressure) • Thermal cycle testing The initial functional testing of the hatch development unit showed no anomalous behavior on the latch trains (all latches functioned smoothly and correctly with the gearbox load inputs ). The engineering team did note the “feel” on the gearbox LRA switching mechanism changed as the system was worn in, but the switc h did successfully move to the armed position and periodic borescope inspections showed the LRA continued to engage as designed. The development hatch was slated for use on the Orion Crew Module Structural Test Article and the hatch delivery was behind schedule, so there was intense programmatic pressure to complete the hatch- level development testing as soon as possible. As a result, the team went into proof pressure testing with minimal instrumentation and no cameras (believing this test would be easily completed because of the Apollo similarity plus the successful subcomponent testing and analysis ). Due to the schedule constraints, t he test team did NOT independently verify the LRA engagement via borescope for this first test and relied completely on the LRA switch being in the armed position. Strain gauges were placed on key areas of all latches and a single LVDT was placed on the exterior of the hatch structure. Figure 12 shows the proof pressure pre- test configuration. The target proof pressure in the test was 161 kPa (1.5 times the crew cab in nominal pressure for the appropriate margin). As the test team slowly increased pressure in stages to 21 kPa, 69 kPa, and 109 kPa, there were no indications of issues . However, at ~90% of proof pressure, a loud noise was heard from the containment chamb er. There were several small windows looking into the room, but no test team members were looking at the hatch directly when the noise was heard. The pressure system was immediately shut down. When the room was opened, the hatch had been completely “libera ted” from the proof pressure fixture (Figure 13 ).

Document

AMS_2020.pdf

477 Figure 12. EDU Hatch Proof Pressure Pre- test Configuration Figure 13. EDU Hatch Proof Pressure Post -test Configuration

Document

AMS_2020.pdf

478 Development Test Failure Investigation The investigation began at once, starting with a fishbone failure diagram. The team arrived at more than 50 possible root cause contributors . An initial look at the latch strain gauges did show non- linearity near the point of failure (see Figure 14 ). The latches shouldn’t behave this way for two key reasons, namely that they are supposed to act as over -center mechanisms, and that the engaged LRA should be a backup structural path to prevent latches from backdriving. These were the leading root cause theories, but it was evident t hat more analysis was needed along with additional testing with improved instrumentation. Figure 14 . B-Train Latch Strain Gauge Data f rom Test Failure The originally planned flex -body analysis was given new urgency. Numerous meetings were held with both NASA and L ockheed Martin expert analysts, and several models were built using various software tools (MSC ADAMS , MSC NASTRAN, ABAQUS). The best (although limited) results came from the ADAMS model (Figure 15) . In this case, the CAD model was used to generate a complete 17 latch kinematic assembly. Flexibility was added to key members that reacted high loads through the use of 1- D springs where stiffness values were analytically derived from the FEM. Latches were modeled using a bushing element at the r oller contact region. Latch stiffness values were derived from the latch subcomponent testing. Finally, the structure was modeled using FEM derived enforced displacements at each latch location. This model was continually updated as the development testing continued, however full correlation was never achieved. The model could show general trends but it wasn’t able to duplicate specific loads or deflections of the latch trains as seen in test ing. This was due to the hundreds of parts and joints, each with different friction and stiffness values. When combined with the complex pressure- induced deflections of the hatch structure, it became clear that the model would be of limited use. Therefore, the team continued to focus on an empirical approach.

Document

AMS_2020.pdf

479 Figure 15 . ADAMS Model of Hatch Mechanisms A design of experiments study was implemented to isolate all variables which resulted in nine additional pressure tests. Many more strain gauges were added to the prototype, particularly in the gearbox, on the linkages connecting the latches, and on the hatch structure. More LVDTs were added plus numerous GoPro cameras inside and outside the pressure vessel to monitor latch movements. The first retest was intended to be a repeat of the initial test conditions, but with slower pressure application and more careful monitoring of all instrumentation. The gearbox LRA was confirmed to be engaged via borescope for this first ret est. Interestingly, the hatch reached proof pressure and the latch non- linearities didn’t recur. This immediately suggested that the LRA had not been engaged for the first test, despite the LRA switch being in the correct position. A second retest was performed, this time with the LRA purposely in the disengaged position. The latch strain gauges showed the same non- linearity as in the initial test. Furthermore, the hardstop gaps on most latches increased as the pressure increased (indicating the lat ches were not acting as traditional overcenter mechanisms). Later tests focused on the rigging procedure for the latch positions. In the original failed test, small gaps were allowed at the latch hardstops (again with the belief that pressure should onl y drive the latches into the hard stops). The rigging procedure was updated to reduce and mostly eliminate the hard stop gaps. Subsequent proof pressure testing with the improved rigging was successful, even with the gearbox LRA purposely disengaged, which demonstrated the LRA could still function as a redundant hatch retention feature. Both the analytical model and the detailed test instrumenta tion supported a fundamental conclusion that as pressure increases on the hatch, it bows the flat structure like a balloon which caused the distance between latches to decrease. Since the linkages connecting the latches are rigid, this defeated the over -center mechanisms, allowing them to release (see Figure 16 ).

Document

AMS_2020.pdf

480 Figure 16 . Latch Movement Due to Hatch Pressure It was also noted the test configuration had several components purposely left out as they weren’t deemed relevant (hinges, thermal isolator brackets, and counterbalance guiderail parts were not installed). These components slightly stiffen the hatch structure and would have decreased pressure deflections (following the Test Like You Fly philosophy is important). At the conclusion of the investigation the root cause was determined to have 3 primary contributors: 1. Hatch structural deformations under pressure- induced latch and linkage loads that counteracted the latch overcenter loads, which caused the latches to move back “undercenter.” 2. Inadequate latch train rigging 3. The gearbox LRA was not engaged despite the LRA switch appearing to function correctl y. This was due to a design issue found within the switch mechanism. Gearbox LRA Switch Mechanism In hindsight, it was fortuitous that the LRA switch wasn’t functioning properly. If it had functioned, the LRA would have prevented the latch trains from backdriving, thus giving false confidence in the latch train rigging procedure and essentially eliminating fault tolerance of the hatch retention system. Nevertheless, it was necessary to understand root cause of the LRA switch issue and resolve it. It is important to note that the LRA is not directly coupled to the Arm/Disarm S witch. Therefore, the LRA could engage without noticeable motion of the switch due to complicated cam features inside the ELAM not operating as intended. Corrective actions to address these issues were implemented in the A rtemis 1 design. The switch functionality did not change except that the Neutral state was removed as it was not needed and only added confusion. The main improvement came by moving the arm/disarm mechanism from the ELAM and into the Gearbox Lid. This allowed for a direct coupling between the LRA and a n added ‘Status Indicator’ that was used to show engagement status. The status indicator displays green when the LRA is engaged and red when disarmed (see Figure 1 7). Table 2 summarizes the relationship between the switch state, status indicator, and LRA condition. It is important to note that this improvement can only be verified by personnel inside the Crew Module. Since the Artemis 1 mission is uncrewed, an additional layer of redundanc y was implemented via a lockout pin installed from the outside of the closed hatch which guarantees a locked position of the LRA. Further improvements for LRA status ha ve been implemented in the A rtemis 2 design to give grou nd crew and astronauts definitive feedback on proper engagement.

Document

AMS_2020.pdf

481 Table 2. Artemis 1 Gearbox LRA Switch States Switch State Status Indicator LRA Condition “Arm” Green Engaged “Arm” Red Ready -to-engage “Disarm” Red Disengaged Figure 17. Improved Artemis 1 LRA Switch Design Artemis 1 Side Hatch Due to the time required to troubleshoot the proof -pressure anomaly, there was insufficient time to complete the hatch development test program. The Artemis 1 flight hatch was already well into manufacturing and assembly while the pressure anomaly investigation was concluding, so it was determined that corrective actions from the anomaly investigation would be implemented immediately on the flight hardware. The improvements to the gearbox and rigging procedure were completed and in late 2018 the Artemis 1 s ide hatch successfully passed proof -pressure testing and a combined pressure/vibration test (a protoflight approach was utilized as there was insufficient time and resources for a dedicated qualification unit) . All instrumentation results from these tests met the pass -fail criteria established from the development test anomaly investigation. The hatch has been successfully integrated on the Artemis 1 vehicle and thermal vac testing of the hatch is being conducted as part of the spacecraft level testing at t he time of writing this paper.

Document

AMS_2020.pdf

482 Lessons Learned Specific lessons learned (and relearned) during the Orion Side Hatch development : • Accurate modeling of complex mechanisms with many parts (especially where flex -body effects exist) may not be possible or feasible. Use analysis results cautiously and t est early. • Beware of using pre- existing designs without thoroughly understanding the effects of your requirements (the Apollo- like latch trains were more sensitive to pressure- deflections from the new flat Orion hatch structure). Additional independent analysis in the early design stages could have identified the pressure deflection failure mode. • Think ahead to possible failure scenarios during testing and ensure that sufficient instrumentation is used to h elp identify and characterize the failure modes should they occur. Still/video cameras should be used even for simple structural tests. • For complex mechanical systems, beware that successful subcomponent tests can impart false confidence for higher level assembly tests. • Ensure that rigging (adjustment) procedures for complex assemblies are well documented with clear pass/fail criteria on critical dimensions and features. • Ensure that the method for verifying the state of mission- critical mechanical functions is reliable (the gearbox latch restraint arm state was not properly established during hatch development testing). Conclusion The Orion side hatch completed a development test program that identified deficiencies in the design which have been corrected for the first uncrewed Artemis 1 mission. The flat structure design (originally thought to be an improvement over the Apollo curved structure design for seal leak rates and latch mechanism performance) introduced new pressure- deflection problems that would have reduced system reliability without correction. The gearbox inspection and hatch rigging procedures were updated which resulted in a successful completion of the Artemis 1 side hatch protoflight testing, and successful delivery/integration into the spacecraft. Acknowledgements • Lockheed Martin Orion Me chanisms team o Paige Carr, Bill Chastain, James Beat, John Lawlor, Doug Harrison • NASA Orion Mechanisms Team o Joseph Anderson, Brent Evernden, Oscar Guzman References • Evernden, B, Guzman, O., “ ORION - Crew Module Side Hatch: Proof Pressure Test Anomaly Investigation”, NASA Technical Reports Server (2018) • Walkover, L, Hart, R, Zosky, E, “The Apollo Command Module Side Access Hatch System,” Aerospace Mechanisms Symposium, 4 (1969), 157- 167

Document

AMS_2020.pdf

483 Design, Development, Testing, and Flight of the Crew Dragon Docking System Jaret Matthews , Caitlin Driscoll* , Edward Fouad*, Andrew Welter *, Marc Jar mulowicz** and Jessica Ipnar** Abstract On March 3, 2019 10:51 U TC, the SpaceX Crew Dragon Demo 1 spacecraft successfully docked to the International Space Station (ISS) on the Node 2 Forward International Docking Adapter (IDA) . The SpaceX developed docking system successfully accommodated the misalignments and relative rates between Crew Dragon and IS S, and ultimately attenuated the kinetic energy following contact . This was the first successful docking of a Commercial Crew program vehicle to the International Space Station. The successful Demo 1 flight was the culmination of years of development . This paper describes the SpaceX docking system, and a sample of the lessons learned through simulation, test and inspection. Figure 1: The Demo 1 Crew Dragon System on Approach to ISS Introduction The SpaceX docking system is comprised of several mechanisms . The soft capture system (SCS) allows for Crew Dragon to compensate for misalignments with ISS resulting from navigation and control errors and provides the initial (soft) attachment to the station. The SCS also is responsible for bringing the capsule to rest with respect to ISS, decelerating it from an initial approach speed at contact with ISS of 8 cm/s. The SCS consists of a ring with three coarse alignment petals, each with embedded passive soft capture latches . The SC S petals and passive soft capture latches are derived from the NASA Docking System design. * Space Exploration Technologies, Hawthorne, CA; caitlin .driscoll @spacex.com ** Dynamic Concepts, I nc., Huntsville, AL; mjarmulowicz@dynamic -concepts.com Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center , 2020

Document

AMS_2020.pdf

484 Figure 2: The SpaceX Docking System conducting a joint interface acceptanc e test; soft capture ring shown in the deployed state The SCS is deployed forward of Crew Dragon via a hexapod – a minimally constrained, six -legged arrangement of two- force struts. As the soft capture ring/petal elements make contact with the IDA, forces travel down the rigid struts and are converted into torques by a moment arm connected to a passive rotary spring damper , also referred to as attenuator arm. The rotary spring damper reacts the contact torque with caged helical compression springs and a rotary eddy current damper. The soft capture system passively attenuates the contact loads and allows the soft capture ring to move in six degrees of freedom (6- DoF), permitting it to align with the passive IDA interface. Assuming sufficient alignment is achieved, the soft capture latch pawls jump over corresponding striker plates on the IDA and soft capture is achieved. Two plungers on each soft capture petals are depressed when capture is achieved, triggering a change in microswitch state. If capture is not achieved, Crew Dragon will perform its failed capture response and safely retreat from the ISS . Following successful soft capture and cessation of relative motion, the soft capture ring is retracted via three linear dual -wound actuators. As the linear actuators are driven up they cause the rotary spring damper s to rotate down and thereby return the soft capture ring to the stowed height. Following soft capture ring retraction, twelve hard capture hooks , derived from the NASA Docking System hook design, connect to corresponding passive hooks on the IDA. The hard capture hooks each contain a cam mechanism that draws the hooks down and compresses a seal that runs around the circumference of the docking system. Ready to hook switches provide the indication that the hooks are in a position to successfully latch onto the IDA passive hooks. Finally, two umbilical dual-wound actuators mate power and data connectors to ISS. Following a successful hard mate operation, the vestibule between Dragon and ISS is pressurized and the hatches opened to allow for transfer of crew and cargo. To undock, the passive capture latches are moded to a “ready to release” state such that they no longer provide load retention capability . The umbilicals are retracted and the hard capture hooks opened, freeing Dragon from the ISS and allowing GNC to take control to move Dragon away from the ISS . Contingency release mechanisms, Non- Explosive Actuators (NEAs) on the soft capture latches and F rangibolts on the hard capture hooks, are present in the event that release via the nominal actuators fails . Notably, the hard capture hooks are also utilized to retain the nosecone during launch and re- entry. This protects docking system and other nearby components for reuse. The SpaceX docking system is a highly performing and efficient spacecraft mechanism. Its weight is approximately 4% of the Dragon 2 mass. It has low power draw and was designed and qualified for fiveflight reusability with minimal maintenance.

Document

AMS_2020.pdf

485 The SpaceX docking system is a well understood mechanism, which has been computer modeled, simulated, and test correlated to a high degree. The system has been modeled in NASA’s Trick software framework, so that this simulation can easily interface with other NASA docking simulations . At many levels of individual components (springs/dampers), sub- systems (soft capture latches, hard capture hooks, rotary attenuator arms), and full docking syst ems, the simulation has been correlated to match test data from both SpaceX and NASA testing. The correlation of the simulation extends to a wide variety of mass properties in the Crew Dragon class, at a wide range of contact conditions and velocities. Figure 3: The SpaceX Docking System Operations Overview Figure 4: SpaceX Docking System Subassembly Overview

Document

AMS_2020.pdf

486 SpaceX Development Approach The SpaceX team has taken a very integrated, systems -based approach to the development of the SpaceX docking system . This integrated approach has been a cornerstone through early concept development, qualification, and flight integration, build and test . Starting with the early concept development, close interaction between GNC and the design engineering teams ensure d accurate system requirements for misalignments and speeds and informed design selection. During the prototype phase, the mechanisms, software and avionics teams worked tightly together to identify the optimum systems design. A system -level qualif ication campaign was undert aken as early as possible to inform docking performance parameters . The SpaceX docking system has undergone extensive qualification testing at both the system and subsystem level, utilizing SMC -S-016 and NASA -5017 mechanisms requ irements . The system as a whole completed shock testing, vibration testing, thermal -vacuum and static structural testing. These system level tests included full system performance tests at thermal extremes. This system level qualification testing highlight ed interactions that may have been missed with solely subassembly qualifications, such as dynamic harnessing effects . This testing campaign allowed the validation and refinement of flight software throughout the testing process . In addition, the soft capture system was subjected to extensive performance testing on the Six-Degree- ofFreedom Dynamic Test System (SDTS) at the NASA Johnson Space Center (JSC). The SDTS was used to simulate the relative dynamics of Crew Dragon docking t o ISS and was critical to understanding system level performance and interactions. Through two testing campaigns, the SpaceX docking system completed more than 450 soft capture events . Significantly, the system and subassembly qualification testing allowed validation of the docking dynamics simulation. Throughout the first and now subsequent flight builds, the close interaction between the integrated teams at SpaceX, as well as dynamics modeling team, has been maintained to ens ure each docking system is built and validated to system requirements . Sub-System Modeling, Correlation, and Lessons Learned The SpaceX Docking System is modeled in NASA /JSC ’s Trick environment . This docking dynamics model was developed by Dynamic Concepts Inc. as a contractor of SpaceX, and highly tuned via subassembly and system level testing. Simulations were created for docking events as well as sub- scale testbeds, which allows for highly tuned as -built data. This simulation was then able to correlate to Demo 1 docking. Sub-System Lessons Learned – Attenuator Arm Acceptance Testing In general, while the attenuator arm is conceptually simple with two springs and a damper, the complexity of mo deling comes from non- linear effects . For the modeling of initial response and peak loads and dynamics, most components can be modeled linearly (all mechanisms have some degree of non- linearity) . However, effects like dead bands, frictions, gaps, local sti ffnesses, all have a major role in accurately simulating the behavior of the attenuator arms. In particular, t he correlation of the attenuator arm simulation resulted in several findings regarding the carriage and the damper. Initial modeling of the deploy structure, or carriage, assumed it was static device, but through correlation with ATP test data and additional testing specific to the carriage, it was found that the carriage not only has significant dead band but also has important back structure stif fness . These effects are critical to model properly due to the high stiffness of the rebound spring. The nature of the eddy current damper, with its high gearing reduction results i s a slight dead band when velocity reverses . This is a minor effect but bec omes more significant at the end of attenuation when the relative velocities between the Dragon and the ISS are small . An example of the correlation in the forces in the strut connecting linkage between the acceptance test and simulation is shown in Figure 5.

Document

AMS_2020.pdf

487 Figure 5: Attenuator Arm ATP Correlation, Test Flight Example Arm Data vs. Simulation Sub-System Lessons Learned – Latch Acceptance Testing The soft capture passive latch model was derived from a high- fidelity dynamics model, based upon CAD geometry of the latch hardware. The Trick model of the latch did not attempt to simulate every possible degree of freedom but used lookup table data to emulate the nominal behavior of the latch when compressing during docking. The latch model included effects due to the unique geometry of the striker, which is not smooth, and produces force spikes during nominal docking. These effects can be seen below in Figure 6. An important lesson learned was the effect that the significant striker geometry had on l atching dynamics, as well as the overall complexity of the latch design and dynamics. Figure 6: Latch Forces vs. Time, Latch ATP Test data and Simulation data System Modeling, Correlation, and Lessons Learned System Modeling and Correlation The Sub- System ATP testing was critical to successfully modeling the SpaceX Docking System behavior when tested at the SDTS facility at JSC . The Six Degree of Freedom Test System facility at NASA’s Johnson Space Center enables soft capture, attenuation, and retraction testing of docking systems in a controlled environment .

Document

AMS_2020.pdf

488 Correlation of the simulation to SDTS testing could only be accomplished due to integration of the simulation with JSC’s simulation of the facility . The facility has several nuances, including controller and actuator effects, which when accounted for, vastly improve the accuracy of the docking simulation. A comparison of the measured forces at the docking plane and the relative vehicle motion, for a representative docking condition, is s hown in Figure 7. Figure 7: Simulation correlation to SDTS test data, Forces and Displacements vs. Time Two system test campaigns were performed on the docking system, an initial campaign in 2015 and a second campaign in 2018 with the final flight de sign iteration implemented. During the first campaign, we learned that the sub- system modeling of the attenuator arms and latches were inadequate. While they could capture the overall behavior of the docking system, there were too many cases which did not correlate properly, many of which were directly due to properly modeling the non- linear behaviors of the attenuator arms . There were also missed captures, which were directly due to improper tolerances of both the latch and striker hardware, but this promp ted the creation of latch models of higher fidelity in order to understand the latch behavior. Several lessons learned came from the model correlation during the second test campaign . Significant ly more sub- scale testing was performed on docking system components, which allowed exact spring rates, preloads and damping rates to be modeled and thus leaving only system level variables as unknowns. This higher fidelity modeling of the attenuator arms resulted in much better correlation to a wide verity of test cases, including those that specifically did not correlate as well during the first test campaign. A major lesson learned was that misalignments of the Soft Capture System ring at deployment do not drive docking dynamics . Due to the passive nature of the docking system, and the relatively small pitch and yaw inertias of the Dragon 2, significant misalignments of the soft capture r ing at deployment were negligible and did not have a strong impact on capture success . This highlights that there is little benefit to high levels of accuracy in SCS ring position, which can drive complexity for systems with Linear Electro- Mechanical Actuators, at the kinetic energies seen in Dragon 2 docking. A critical finding from testing was that the capture switches could be triggered or cycled while the mechanism was captured, under heavy shear/bending load. This effect could be replicated with the simulation, after updates to the simulation using a more advanced capture switch model . The effect results from an interaction between the specified free play between the latches and strikers, in addition to the location of the capture switches, nominally on the edge of the radius of the passive soft capture system ring. This led to an increase in a flight software persistence value, specifically the number of consecutive

Document

AMS_2020.pdf

489 control cycles where the capture switch is not pressed to initiate the failed capture response, to account for this affect and prevent false negatives during the mission. The high accuracy of the simulation meant that properly modeling the gravity off -loader became significant . To better replicate the on- orbit dynamics, the SCS ring has an off -loader which reacts the force of gravity from the SCS ring. Howev er, this somewhat simple device can have significant frictions and stictions which become difficult to model for each unique run. A lesson learned from SDTS testing is to improve the repeatability of the gravity off -loader and fully characterize its behavi or to aid in model correlation. Flight Reconstruction The Demo 1 test flight mission was reconstructed with the docking simulation and the flight telemetry. The simulation had good correlation to the flight contact and attenuation dynamics, as well as the retraction and hard capture dynamics . Figure 8 compares the Flight Telemetry of the Attenuator Arm Encoders to the simulated encoders for capture. Figure 8: Flight Attenuator Arm Telemetry vs. Simulation vs. Time For undocking, there were two principal sources used for correlation, the Ready to Hook switches, and the calculated Dragon velocity . The ready to hook switch states were used to verify the hook dynamics and overall Dragon separation state. The calculated Dragon velocity was used to ve rify the forces separating the Dragon from the ISS . Notably, there appeared to be little to no seal stiction (expected given the se al lubrication), and the latches provided nearly no restraining force, resulting in significant departure velocity when the hard capture hooks cleared. Properly modeling the hooks and their motion, in order to determine when there is clearance for departure, is a critical lesson to modeling undocking. Qualification Testing Failures and Lessons Learned There were many lessons learned across the various test campaigns, several of which manifested as test failures . These early testing lessons learned led to design updates and delta qualification testing before the successful Demo 1 flight test and ultimately a more reliable system . A subset of these failures and their corrective actions are outlined below.

Document

AMS_2020.pdf

490 Figure 9: (Left): The SpaceX undergoes dynamic testing on the NAS A JSC Six -Degree- of-Freedom Dynamic Test System. (Right): The SpaceX Docking System prepared to enter thermal vacuum testing to perform system level qualification tests. TVAC Harnessing Failure During system level thermal vacuum testing, there was a loss of signal from the controllers during the soft capture deploy sequence. To troubleshoot, the vacuum chamber was opened, and it was determined that the connectors to the soft capture latch controllers had sheared. Subscale testing was performed, and it was determined that the ov erwrap material of the harness, a rubber -like material , was stiffening the harness in vacuum, which caused it to shear. By replacing this overwrap material with a more flexible material , this condition was eliminated. This provides a valuable lesson learned in harnessing material selection, especially in dynamic mechanisms operating at a wide temperature range. Figure 10: Harness Connector Sheared during Deployment

Document

AMS_2020.pdf

491 Microswitch Testing Defects Microswitch es are used throughout the SpaceX docking system as position indicators. In addition to the soft capture petal switches, which are used to declare capture, and the ready to hook switches, which indicate when the IDA hard capture plane has been contacted, s witches are used within the attenuator arms and the hard capture hooks to indicate the mechanism state. Durin g system and sub- scale testing, microswitch failures were observed. Investigation revealed fracture of internal components to the microswitch caused during shock events. These failures lead to the selection of a different series of microswitches with fully potted internals that were more robust to shock events. This was an important lesson learned in assessing the robustness of commercial of the shelf components to shock induced failures. Additionally, it was found that the harness stiffness was such that was subject to fatigue failure modes, even from the most careful of handling. A less stiff harness material was used to prevent this failure mode. This was an important lesson in selecting harness potting materials that are robust to nominal handling. Figure 11: Fracture of Ceramic Contact Pin in Shock (left) and undamaged switch (center) ; harness damage (right) Inspection Lessons Learned In addition to thoroughly testing hardware, the SpaceX docking system has gone through intensive inspection campaigns . Visual inspections and tear downs are a critical component of qualification campaign. SpaceX has also greatly increased its use of Computerized Tomography (CT) inspections. This allows hardware internals to be inspected without a tear down that inherently changes the condition. CT inspection of hardware has proven to be an immense resource for both failure investigations as well as identifying potential failure modes . Two examples of SpaceX’s use of CT scanning are outlined. It is important to note that CT scans do provide radiation dosage to active components. A dosimeter scanned with a development unit can estimate this dosage to determine impact on the unit’s usable life. Retaining Ring Vendor Verification After qualification testing, the eddy current damper was CT scanned, though it performed nominally during the campaign. The scan revealed a significant defect : a reta ining ring had become dislodged from i ts groove. If settled in a slightly different position, t his retaining ring could become stuck in the gearbox, causing the damper to jam. This led to additional development testing, where it was discovered that only improperly seated retaining rings were at risk of becoming dislodged; properly seated retaining rings survived qualification environments as indicated by pre- and-post environment CT scanning. Through this discovery, SpaceX worked with the vendor to implement robust quality control practices on retaining ring install. CT scan videos are stored for future evaluation, which proves useful if a different failure mode was discovered: videos can be reviewed without requiring hardware to be removed and re- scanned. This was an example where CT scanning enabled discovery of a potential problem without experiencing a dramatic

Document

AMS_2020.pdf

492 failure; CT scann ing of both vendor and internal assembled components can be utilized as a very informative and robust quality control tool . Figure 12: Dislodged Retaining Ring Harness Failure During an acceptance test, an intermittent connection was noted on one of the power and data buses. The problem was isolated to one connection. The controller was removed and tested and no anomalies noted, which led to additional investigation of the harnessing. It was found that ‘wiggling’ the harness could cause the bus to disconnect . The harness was CT scanned to identify the root cause: improper insulation on one wire allowed the cabling to ‘bird cage’, which could then short against the connector . The insulation was not properly adhering to the cable, which would allow the insulation to pull off the cable and induce this condition. CT scanning provides extremely clear indications of this root cause. This issue could have been easily written off as a single workmanship issue; however, through CT scan it was identified that this was not an isolated incident but cable adhesion suscept ibility; and through further testing the faulty cable lots were identified and contained. This is another example where CT scanning can be a valuable investigation tool when failures occur. Figure 13: CT Scan images showing ‘bird caging’ of cable which had its insulation improperly adhered and pulled up. This allowed the cable to short against the connector housing

Document

AMS_2020.pdf

493 Lessons Learned: Systems Interactions SpaceX benefits from having a highly integrated team of software, avionics, operations, GNC and mechanisms engineers. This integration is critical to the successful design and continued operation of the SpaceX docking system. The close interactions between flight software and the mechanical team, for example, allowed for very robust testing of both hardware and software: the flight software algorithms, fault detection methods, voting schemes and responses were used for all system level testing. This allowed early validation that the software responded as expected, includi ng in off -nominal cases . This close interaction also allowed validation of configurable parameters, and the accurate modeling of flight software response in simulation allowed small changes in these parameters to be validated. For example, a key parameter in the docking configuration is the persistence value to declare soft capture and loss of capture from the soft capture switches . By using flight software during 6- DoF testing at JSC, and then adjusting this configurable value in the docking simulation, this parameter could be fine -tuned for the desired response . The close interaction between the simulation and GNC allowed rapid iteration based on expected contact conditions and docking parameters. The integration between avionics teams and hardware teams was also critical to the rapid and successful development of the SpaceX docking system. Controllers were developed specifically for the docking system application, and close coordination during subassembly and system level testing allowed for the fine- tuning of controller firmware. With a dynamic system, harness routing to the soft capture latches and switches is critical; a misrouted harness could impede motion and have drastic impacts on mechanism success. Close coordination with harnessing engineers, inc luding involving them in system level testing so they could better understand the expected range of motion, allowed for a successful harness design. Conclusions The SpaceX docking system is a custom implementation of the N ASA Docking System architecture . It is the first American designed and manufactured docking system to be flight proven in decades . The SpaceX docking system leverages a semi -passive system to effect docking capture, attenuation, and alignment before retraction to hard captur e. It is effective for a broad range of contact conditions, is low power , light weight, and cost effective. The system has been extensively simulated, tested, and inspected to ensure successful operation throughout a wide range of possible contact conditions. Figure 13: The Demo 1 Crew Dragon Capsule Docked to the ISS Forward IDA in March 2019.

Document

AMS_2020.pdf

494 References 1. McFatter, J., K. Keiser, and T. Rupp , “NASA Docking System Block 1: NASA’s New Direct Electric Docking System Supporting ISS and Future Human Space Exploration” Proceedings of the 44th Aerospace Mechanisms Symposium, NASA Glenn Research Center, May 16- 18, 2018 2. NASA Docking System Interface Definitions Document (IDD), JSC 65795, Revision J

Document

AMS_2020.pdf

495 Approved for public release. OTR 2020- 00294 Highlights of t he Next -Generation AIAA Moving Mechanical Assemblies Standard Brian W. Gore* and Leon Gurevich* Abstract Over the past several years, an effort has been under way to transform the previously rele ased AIAA S-114 (2005) standard (Moving Mechanical Assemblies for Space and Launch Vehicles) into a more modern set of requirements to complement the way many space acquisition programs have evolved. A joint government /industry committee of mechanisms -minded subject matter experts was formed, and the previous standard was reviewed cover -to-cover by that committee for opportunities to make it reflect modern- day philosophies. The actual standard, tentatively designated AIAA S -114A, has not yet been released as of the publication of this paper, since it is currently making its way through final formatting and the approval process, but this paper discusses the most significant changes in a general way as a preview of what will be seen in the final release. Introduction 15 years…that is a long time! Uber and Lyft did not exist to deliver us door -to-door yet . Instagram and Snapchat had not yet taken over our teenagers’ lives . Even the ubiquitous iPad was still 5 years away! But that is how long it has been si nce the AIAA standard “Moving Mechanical Assemblies for Space and Launch Vehicles” was first released. New technologies and products, better methodologies, and smarter, more innovative ways of performing testing and analysis have all come along in that tim e, creating the need for a more up- to-date set of requirements to design, build, and test moving mechanical assemblies. This paper includes a qualitative description of the update process and the more significant changes that resulted and will be found in the new Moving Mechanical Assemblies (MMA) standard, tentatively and likely designated as AIAA S -114A. After concur rence was received from AIAA to move forward with the update , the authors sent ou t invitations to government and industry stakeholders in the mechanisms expertise area to participate on a committee of such subject matter experts . That committee’s charter was to approve by consensus the changes best suited for incorporation into the updated standard, such that it would be the best compromise of cost -effective requirements and guidelines to produce robust MMAs for space and launch vehicle applications. Much of the MMA standard’s material in the form of requirements has been retained sinc e there was already significant sound work done by the previous committee that produced the 2005 version of the standard. However, that committee did run short on time to address some shortcomings, specifically in the testing section. As a result, the test ing section was a top priority in this update effort . It has been reformatted and rewritten, with a goal of emulat ing the format of SMC -S-016 (also TR-RS-2014- 00016, or formerly MIL-STD-1540 C), “Test Requirements for Launch, Upper Stage, and Space Vehicles” (References 1, 2, and 3) . Moreover, a completely new approach to calculating force/torque margin is presented in this new revision. It specifies reduced margins over certain characteristic resistances that are more predictable or repeatable, but maintains the original margins for variable resistances such as friction . A detailed explanation of the process, and even a working example, are included in the new version of the standard. * The Aerospace Corporation, El Segundo, CA; [email protected]; [email protected] Proceedings of the 45th Aerospace Mechanisms Symposium, NASA Johnson Space Center, 20 20

Document

AMS_2020.pdf

496 Approved for public release. OTR 2020- 00294 Perhaps the next most significant change that users of the new MMA standard will notice is the elimination of the “shall, where practical” weighting level of requirements. These requirements were deemed to be unverifiable and thus , difficult to enforce. A concerted parsing effort was launched to identify each of these approximately 90 instances in the standard, and to either upgrade each occurrence to a “shall” weighting level requirement, or downgrade it to a “should” weighting level for best practice or guidance. An additional goal of this parsing effort was to ident ify all of the requirements that were vaguely worded or non- verifiable. For example, approximately 40 additional requirements that used terms such as “shall be considered” or “shall be minimized” were found and corrected or modified . Overall, more than 350 comments and requests for revisions were compiled by the government and industry committee working on this revision. Each comment was discussed and adjudicated by consensus . The comments fell within three broad categories: testing related , non- testing related , or parsing to eliminate all instances of the “shall, where practical” weighting level and other vague requirements . This paper will summarize the more significant changes that are expected to be published in AIAA S -114A. The MMA Update Process and Industry Committee Starting at a low -level of effort in 2014, the authors of this paper reviewed the material in the previous standard that had been repeatedly identified as credibly contentious, outdated, or troublesome, based on repeated mutual experience in program tailoring efforts , hardware testing campaigns, and day -to-day discussions and working- level meetings with customers and space and launch vehicle organizations . These items were categorized into three areas: • The “shall, where practical” weighting factor • The clarity , sequence/flow, and completeness of the testing section • Other items identified as needing timel y updates or matur ation , based on real -life experience since the last standard’s release All items of all categories were compiled into a comment resolution matrix (CRM) in a spr eadsheet format (see Figure 1 for sample excerpt) in order to track what the original and proposed change to the text was, and the rationale for changing it . Figure 1. Sample CRM entries The authors proposed many of the initial changes in the text of the new draft MMA standard document , where l ittle or no contention among stakeholders or user s of the standard was expected. Some examples of these were typographical and grammatical errors, practices that were no longer supported or followed by the bulk of the industry, etc . Other, more significant , proposed changes were left open- ended in the CRM for future resolution and adjudication by an indus try committee of mechanisms subject matter experts. Comment AuthorOrganization Section SubjectComment TypeExisting Content /text Proposed content/text Rationale Committee Discussion Comments Committee DispositionVerified as Implemente d? Comments and Recommended Changes BoesigerLockheed Martin Space8.1.2 Technical at the vehicle test level at the vehicle test level if accessiblecannot see them all at vehicle level Accepted Yes Brandan RobertsonNASA8.1.2.c Typical inspection elementsTechnical fastener torque - torque on pre-selected fasteners before and after exposure to each environmental test conditionfastener torque - torque on preselected fasteners or a check of torque striping before and after exposure to each environmental test conditionTorque striping is a simple, inexpensive, and effective check on fastener loosening that is for some reason often overlooked.Consider incorporating this into re-write as described in comment 29. (maybe add "where used") See proposed language in comment 89Accepted with Changes Yes D. SheltonLockheed Martin Space8.2Test Instrumentat ionTechnicalTest instrumentation should be calibrated in place through the test electronics and in the environmental conditions of the test.DeleteTest instrumentation may be fully calibrated or may be reference only for a development test. Calibration should be covered in general up front in the document.Test instrumentation SHALL be calibrated consisent with the environmental conditions of the test.Accepted with Changes Yes Mark BalzerJPL8.2 Torque checksTechnical ? Duplicate of other comments (e.g. 30) Rejected Yes

Document

AMS_2020.pdf