Solar Dynamics Observatory Reaction Wheel Bearing Friction Increase: Detection, Analysis, and Impacts

Transcription

1 SpaceOps Conferences 5-9 May 2014, Pasadena, CA SpaceOps 2014 Conference / Solar Dynamics Observatory Reaction Wheel Bearing Friction Increase: Detection, Analysis, and Impacts F. Matthew Ekinci 1 Honeywell Technology Solutions, Inc., Greenbelt, MD, The Solar Dynamics Observatory is a three-axis stabilized NASA heliophysics satellite operating in geosynchronous orbit utilizing reaction wheels as its primary attitude actuators. Beginning in April 2013, the SDO operations and instrument teams began observing an increase in spacecraft jitter that was ultimately traced back to anomalous behavior in one of its reaction wheels believed to be related to fluctuations in bearing friction. Though the problem seemed to resolve itself over the course of several months without any corrective action, on February 16, 2014 the same reaction wheel became temporarily stuck at zero RPM and required a torque significantly higher than nominal to resume spinning, inducing an attitude error of several arcseconds in the spacecraft when it did so. Asessment of the data from the incident by the manufacturer has suggested that the wheel is healthy and drastic mitigations such as resting the wheel are unnecessary, but that extended periods of operation in an undesirable low-speed range should be avoided in the future. Options under consideration to ensure this include modifying the SDO momentum unload scheduling process so as to avoid low speeds and utilizing a flight software command to rapidly force the wheels through the undesirable range. I. Introduction AUNCHED on February 11, 2010, the Solar LDynamics Observatory (SDO) is a three-axis stabilized NASA heliophysics satellite operating in geosynchronous orbit. SDO carries three sun-observing instruments: the Atmospheric Imaging Assembly (AIA); the Helioseismic and Magnetic Imager (HMI); and the Extreme Ultraviolet Variability Experiment (EVE). In its nominal observational Science Mode, SDO continuously maintains a sun-fixed attitude aligned with with Solar North Reference (SNR) frame, Figure 2. SDO reaction wheel configuration. Figure 1. SNR frame representation. which is defined by the vector from the spacecraft to the center of the sun, the cross product of that vector with the solar north pole vector, and the cross product of the other two axes that completes the orthogonal set. A visual representation of the SNR frame can be seen in Figure 1. Each of the four AIA telescopes is mounted with a guide telescope consisting of a series of four photodiodes configured such that sun presence is only maintained on each of the diodes 1 SDO Flight Dynamics Lead, NASA\GSFC Code 444.0, Mail Stop 428.2, AIAA Member. 1 Copyright 2014 by the, Inc. The U.S. Government has a royalty-free license to exercise all rights under the copyright claimed herein for Governmental purposes. All other rights are reserved by the copyright owner.

2 when the spacecraft is aligned with the center of the solar disc. While in Science Mode, attitude determination for the spacecraft s pitch and yaw axes is provided by a designated Controlling Guide Telescope (CGT), while roll knowledge is provided by an Extended Kalman Filter (EKF) utilizing inputs from a sensor suite consisting of two Selex ES star trackers, an Adcole digital sun sensor, and three Kearfott two-axis inertial reference units (IRUs). Error signals from the CGT and Kalman Filter are then used by the SDO Attitude Control System (ACS) flight software to determine appropriate torque commands for a set of four Goodrich Type E reaction wheel assemblies (RWAs) that serve as SDO s Figure SDO hourly-averaged reaction wheel speeds. primary attitude actuators and enable fine pointing control 1. As seen in Fig. 2, SDO s four RWAs are arranged in a symmetric pyramidal configuration about the spacecraft X (roll) axis such that the operation of any three wheels ensures full three-axis control. In order to limit induced spacecraft jitter, the reaction wheels are maintained within an operational limit of +\- 850 RPM. Each of the wheels generally experiences one transit of the majority of that operating range over the course of two to four months as the control system adjusts for the effects of external torques, with momentum unloads via thruster firings scheduled when any wheel is predicted to exceed the operational speed limit within the following week. An example of the evolution of the reaction wheels speeds over time can be seen in Fig. 3, which depicts the hourly-averaged wheel speeds for the entirety of II. Wheel Anomaly Detection The EKF and CGT error signals are generated as part of the 5 Hz ACS operating cycle, and are nominally downlinked as telemetry at a variable lower rate (nominally 1 Hz for the EKF outputs and 0.25 Hz for the CGT error signals while in Science Mode) for bandwidth management purposes. While moment-tomoment monitoring of this telemetry is handled by the SDO ground system and fault detection and correction software aboard the spacecraft, the SDO flight operations team Figure 4. SDO roll, pitch, and yaw attitude error over duration of April 28, (FOT) routinely monitors these and other critical telemetry points for long-term trends and changes in behavior. Beginning on April 2

3 Figure 5. Standard deviation of AIA ISS errors on April 28, , 2013, the FOT observed an increase in the noise of the attitude error signals for the spacecraft roll and pitch (Y) axes, with the standard deviation of the error signals increasing by a factor of approximately three. Figure 4 displays the unaltered EKF-based roll attitude error and CGT error signals for both the pitch and yaw (Z) axes as they were received on the ground that day. In addition to the ACS sensor suite, the AIA and HMI piezoelectric instrument stabilization systems (ISSs) also provide information on jitter by returning error signals based on the perceived motion they remove from the instruments. These data are nominally downlinked once every several seconds, though for diagnostic purposes the rate may be upped in three-minute bursts to 128 Hz for AIA telescopes #1 and #3, 256 Hz for AIA telescopes #2 and #4, and 512 Hz for HMI. On April 29, the AIA and HMI instrument operations teams reported that they had seen similar rises in the baseline noise level of the ISS errors for both instruments during the previous day. Figure 5 displays the standard deviation of the AIA ISS error signals over the duration of April 28, 2013; 50 dn corresponds to approximately 1 arcsecond. Though the instrument teams reported that the magnitude and frequency of the noise was within the removal capabilities of the ISSs (and as such was having no noticeable impact on science collection), this incident still represented a significant deviation from a previously stable long-term behavior. Additionally, while this episode of increased jitter initially appeared similar in signature to other attitude disturbances SDO had experienced over the course of its mission, it eventually became clear that this was a unique situation independent of those prior incidents (which had largely been dismissed as brief excitations of spacecraft structural modes by the reaction wheels). Whereas prior incidents had lasted on the order of hours, the event beginning on April 28 subsequently began to stretch into days and eventually weeks. Furthermore, previous attitude disturbances had presented a bell curve signature of smoothly and symmetrically increasing and decreasing jitter centered around a central peak, whereas this particular incident involved erratic, varying behavior. Figure 6 displays the hourly mean, minimum, and maximum for the CGT error signals from April 19, 2013 through the end of May 30, 2013; the large spike on May 15 corresponds to an SDO thruster maneuver, while the smaller spikes observable in yaw on May 10 are the result of a reaction wheel crossing through zero RPM. Given this unusual behavior, the FOT initiated an investigation into possible root causes. Figure 6. Hourly mean, minimum, and maximum CGT error signals. The magnitude and strong directionality of the jitter almost immediately steered the investigation in the direction of the reaction wheels, as other moving parts on the spacecraft capable of causing such motion were either too small and\or infrequently mobile (e.g., instrument filter wheels and the inertial reference units) or moved freely in multiple planes (e.g., the spacecraft s two high gain antennas) 2. A review of reaction wheel telemetry from the first several weeks following the start of the increased jitter revealed a series of unusual discontinuities and trends in the data from RWA4, one of the two wheels aligned with both the spacecraft roll and pitch axes in which the jitter was being observed (as seen in Fig. 2). Time-averaging of the data to dampen noise proved particularly useful for discerning trends that were not immediately clear to the naked eye in short-term plots of the raw data; Figure 7 shows the

4 Figure 7. Hourly-averaged RWA4 wheel speed, commanded motor torque, bearing temperature, and current draw. hourly-averaged values for RWA4 s speed, commanded torque, bearing temperature, and current draw over the first five months of Step changes in the wheel speed correspond to SDO thruster maneuvers, while the sudden series of temperature fluctuations in March represents one of the semiannual periods during which SDO experiences a daily solar eclipse. III. Initial Trending and Assessment Following the identification of the changes in RWA4 s behavior, an initial discussion was held to assess whether a factor external to the RWA specifically, an ongoing and separate issue with the IRU aligned with the spacecraft X and Y axes, or an unexpected and significant change in environmental torques could be driving the changes by causing the ACS RWA control algorithm to chase noisy attitude and rate error estimates. This notion was quickly dismissed, as the momentum distribution law governing the reaction wheels would have caused a similar effect to be seen in the other wheel aligned with the same axes as RWA4 (RWA2), and no unusual changes had been observed in the IRU telemetry. The focus then shifted to identifying potential issues within RWA4 itself. Trending of the RWA telemetry as a function of wheel speed and\or commanded torque rather than time proved useful during the initial telemetry review for differentiating between non-nominal values and those to be expected based on pre-launch data and on-orbit experience. Given the discrepancy between the magnitudes of the commanded wheel torques (generally on the order of 10-3 N m) and SDO s derived average net environmental torques Figure 8. RWA4 hourly-averaged commanded torque and wheel speed from March 18, 2010 through June 1,

5 (on the order of 10-6 N m), the commanded torque for a given wheel was assumed to be an effective approximation of the drag torque the wheel was experiencing. As such, plots of the reaction wheel commanded torques as a function of wheel speeds were used to establish drag torque profiles for each of the wheels over the life of the mission and to assess the significance of deviations from expected behavior; the validity of this method was affirmed by the RWA manufacturer, who stated that it served as one of their primary methods for assessing bearing friction changes in on-orbit wheels. Figure 8 depicts the hourly-averaged torque-speed curve for RW4 from March 18, 2010 (the day SDO reached its geosynchronous orbit) to June 1, 2013, with each point color-coded to represent the time at which it was taken. As can be seen in Fig. 8, RWA4 began requiring mildly elevated levels of commanded torque for a given wheel speed (corresponding to increased drag) early in 2013, though the problem remained otherwise asymptomatic until the increase in spacecraft jitter and shifts in RWA4 s bearing temperature and motor current draw previously discussed began on April 28. Those changes were found to coincide with the significant increase in wheel drag also observed on that date and visible in Fig. 8 at approximately -220 RPM. Subsequent changes in those parameters were also traceable back to the fluctuating pattern of wheel drag that continued to persist in the following weeks. During an initial meeting on June 4, the manufacturer indicated that the problem was likely due to one of two causes: either bearing retainer instability ( chatter ) caused by changes in viscous friction, or lubricant starvation in the bearing. While the latter represented a more serious condition in the near term, the existing data were inconclusive and mitigating actions put forth by the manufacturer (e.g., resting the wheel by allowing it to spin down to zero RPM and remain there for several weeks) involved Figure 9. RWA4 wheel speed and commanded torque on 5 substantial impact to SDO s operations. Additionally, RWA4 was predicted to pass through both zero RPM and the -220 RPM range June 16, within the following four weeks; these events would help more conclusively determine the root cause of the anomaly, as lubricant starvation would likely manifest itself as a significant increase in breakaway torque during the zero crossing while additional anomalous behavior in the same RPM range it had been observed in previously would support the idea of the bearing retainer going into resonance due to changes in viscous friction. Based on these facts, the manufacturer recommended that the SDO FOT take no immediate action but instead continue to closely monitor and trend the wheel data through the periods of interest. RWA4 passed through zero RPM on June 16, 2013; Figure 9 displays the unaltered RWA4 speed and commanded torque over the duration of that day. While the wheel briefly required a slightly elevated amount of torque to reverse direction and resume spinning, the manufacturer stated that the observed behavior was not consistent with the level of increase expected in the case of lubricant starvation. Other reaction wheel and ACS telemetry was within expected ranges during the event. On June 22, 2013, RWA4 crossed the -200 RPM threshold and again began seeing increases in the wheel drag disproportionate to the change in the wheel s speed, with the commanded torque Figure 10. RWA4 hourly-averaged commanded torque and wheel speed from March 1, 2013 through March 1, peaking at 15 mn m when the wheel reached -400 RPM. However, the unpredictable fluctuations in torque seen earlier in the year in the same wheel speed range were absent. Additionally, once the wheel passed through -400

6 RPM, the wheel drag began to decrease, eventually returning to a range in-family with those seen throughout the rest of the mission following a momentum unload that was conducted on August 21. Figure 10 displays RWA4 s hourly-averaged commanded torque versus hourly-averaged wheel speed from March 1, 2013 through March 1, Figure 11. RWA4 wheel speed and commanded torque on February 16, Figure 12. SDO attitude error on February 16, IV. February 16, 2014 RWA4 Zero Crossing Following the summer of 2013, the FOT continued to closely monitor the RWA4 telemetry trends, but no further anomalous behavior was observed for the duration of 2013 and the first several weeks of 2014 (as can be seen in Fig. 10). On February 16, 2014, RWA4 passed through zero RPM for the first time since June 16 of the previous year. The crossing was accompanied by spikes in the spacecraft attitude error of approximately 12 arcseconds about the roll axis and 7 arcseconds about the pitch axis, a significant enough disturbance to briefly trip the onboard fault detection and correction limit for Science Mode roll axis attitude error and alert the SDO on-call engineer. Further investigation showed that upon RWA4 reaching zero RPM on the morning of the 16 th, the ACS began issuing RWA4 progressively larger torque commands with no response from the wheel until the commanded torque reached a value of approximately 26 mn m; at this point the wheel suddenly began rotating again, and the aforementioned attitude disturbance was induced on the spacecraft. While this level of breakaway torque was unprecedented in the prior on-orbit operation of the wheel, it was still well within the wheel s software-limited maximum torque of 0.25 N m. Figure 11 displays the unaltered RWA4 wheel speed and commanded torque over the course of February 16, while Figure 12 displays the unaltered EKF-based roll attitude error and CGT error signals for both the pitch and yaw axes. Despite the substantial increase in breakaway torque at zero RPM, RWA4 s wheel drag immediately returned to nominal levels in-family with those seen earlier in the mission once the wheel began spinning again, and continued to behave nominally in the following weeks. Following the incident, the manufacturer s assessment of the data concluded that the behavior was unlikely to be caused by bearing lubricant starvation, and more probably related to the extended period of time the wheel had spent in the low-speed regime (categorized as +\- 50 RPM) prior to going through zero RPM. While not an immediate threat to the wheel s health, their recommendation was that the wheel s allowed time noted low-speed region should be minimized in the future in order to avoid similar problems and for the benefit of the long-term performance of the wheel. V. Potential Mitigating Actions Several mitigating actions aimed at ensuring the health of RWA4 and not further exacerbating any possible existing problems have been considered since trouble with the wheel first began in Of these, the most 6

7 disruptive and intrusive to the spacecraft is the previously-mentioned resting of RWA4. While the manufacturer s assessments of the RWA4 data have led to this option being deemed currently unnecessary, it is the primary method for addressing lubrication starvation and severe retainer chatter problems and would be turned to were such situations to develop. In this scenario, the FOT would command the SDO ACS to remove RWA4 from its control loop, eliminating the application of control torques and allowing the wheel to naturally spin down to zero RPM. At that point the wheel would be allowed to sit unused for a period of five to six weeks (as recommended by the manufacturer), over the course of which the remaining bearing lubricant would hopefully redistribute itself so as to reduce or eliminate future similar friction-related problems. This process could then be repeated periodically as necessary. While easy to implement for SDO as it requires a simple uplink of new parameters for a stored flight software table, the removal of RWA4 from the control loop would reduce the spacecraft s effective momentum capacity about its X-axis by 25% and about its pitch axis by 50%, significantly increasing the frequency with which momentum unloads would need to be performed during the resting period. Moving to three-wheel operations would also impose an additional constraint on SDO s semiannual roll calibration for the HMI instrument; without a fourth wheel, there would no longer be a guarantee that the remaining reaction wheels speeds would remain within their operational limits over the course of the calibration (which involves a 360 roll about the spacecraft sunline). A second mitigation option focused on avoiding the minimizing the impact of time spent in the undesirable lowspeed region is to change the process by which SDO s momentum management is planned and conducted. Currently, the exit momentum targets for SDO s momentum unloads are selected to maximize the time between unloads and as a result minimize the amount of cumulative science data lost due to thruster maneuvers; no consideration is given to the wheels speed profile, including zero RPM crossings, so long as excursions outside the 850 RPM limit are avoided. By modifying the target planning to instead concentrate on the avoidance of low wheel speeds and conducting momentum unloads rather than allowing the wheels to pass through zero RPM, problems related to the low-speed regime can therefore be avoided by avoiding the regime itself. As with resting the wheel, the obvious drawback to this option is a significant increase the number of momentum unloads performed per year, though without the additional risks incurred by directly inducing physical changes in the RWA as with the resting option. While conducting additional maneuvers poses no fuel consumption problems for SDO (the spacecraft currently has almost 385 kg of propellant remaining and only uses an average of 40 g per momentum unload), it would drastically increase the amount of science data lost annually due to thruster maneuvers. SDO current performs approximately four momentum unload maneuvers per year, each of which causes the loss of approximately 15 minutes of data. As such, the first two options described could be expected to result in the loss of no less than an additional hour of data per year. A third mitigation option, and one which would avoid data loss, is to move away from the current ACS concept of operations in which SDO s reaction wheel speeds are allowed to evolve naturally based solely on the environmental torques encountered by the spacecraft. As described in References 3 and 4, the SDO ACS supports the application of a commanded bias to the reaction wheels momentum in the wheel frame such that a non-unique set of wheel speeds produces the same total momentum in the spacecraft body frame. This bias command was utilized during SDO s commissioning phase to drive the wheels through wide ranges of speeds as part of on-orbit jitter testing, but has not been used to manipulate the wheel speeds during the mission s science phase. By applying a bias as a given wheel (or specifically RWA4) approaches zero RPM, the ACS could be directed to drive the wheel through or away from zero RPM and eliminate the possibility of it lingering in the undesirable low-speed regime. Once the correction for the effects of external torques on the spacecraft has built the wheel speed to a state where it would remain outside the low-speed regime without the momentum bias, the bias may then be removed in order to reduce the wheels speeds and avoid excessively shortening the timespan between necessary momentum unload maneuvers due to the wheels 850 RPM operational limit. Based on simulations performed using the SDO High-Fidelity Simulator (a Simulink model and simulation platform constructed for use in pre-launch requirements verification and post-launch maneuver simulation) and examination of the ACS telemetry from when the null bias command was utilized during jitter testing, it is expected that the application of the null bias would enable the wheel(s) to be driven through their undesirable range in approximately one minute with minimal (1-2 arcsecond) attitude errors induced on the spacecraft; these errors are sufficiently small so as to be manageable by the ISSs, enabling nominal science operations to continue during the process. However, the operational wheel speed limit also imposes a significant constraint on the use of this mitigation option, as it may not be possible to implement it without violating the limit in the case where one wheel is near the limit as another approaches zero RPM. It would be necessary to perform a momentum unload maneuver in 7

8 such a situation in order to avoid the low-speed regime, though the ability to bias the wheel speeds in either the positive or negative direction would ensure that such cases would seldom occur. VI. Conclusion The anomalies related to RWA4 have provided useful lessons to the SDO project in terms of the importance of cross-functional cooperation for trending, data assessment, and anomaly investigation, while also reinforcing the importance of trending related and\or dependent telemetry parameters as a function of each other rather than strictly time. With six years remaining in SDO s ten-year operational goal and ostensibly extended operations beyond that, the importance of maintaining the health of the actuators that enable the spacecraft s science mission is paramount. As such, while RWA4 currently continues to behave nominally, the FOT will continue to closely monitor the drag behavior of all four reaction wheels and explore operational mitigations and contingencies related to anomalous wheel behavior in order to more rapidly identify developing problems in the future and take action as appropriate. Acknowledgments The author would like to thank Dale Fink, Dean Pesnell, and the Space Science Mission Operations project office for their support of this work. The author would also like to thank Zoe Frank for the provision of the AIA ISS data utilized in Figure 5, and the SDO ACS development team, who originally produced Figures 1 and 2 for internal documentation. References 1 Starin, S. R., Bourkland, K. L., Liu, K.-C., Mason, P. A. C., Vess, M. F., Andrews, S. F., and Morgenstern, W.M., Attitude Control System Design for the Solar Dynamics Observatory, Flight Mechanics Symposium, NASA CP , Liu, K.-C. A., Blaurock, C. A., Bourkland, K. L., Morgenstern, W. M., and Maghami, P. G., Solar Dynamics Observatory On-orbit Jitter Testing, Analysis, and Mitigation Plans, 2011 AIAA Guidance, Navigation, and Control Conference, AIAA, Washington, DC, 8-11 August doi: / Natanson, G. et al, Solar Dynamics Observatory (SDO) Ascent Planning and Momentum Management, Spaceops 2010 Conference, AIAA, Washington, DC, April doi: / Mason, P. A. C., Starin, S. R., SDO Delta H Mode Design and Analysis, 20th International Symposium on Space Flight Dynamics, Annapolis, MD, September