Gearing and Actuators

Speed reducers and complete actuators selected by flown and qualified planetary and orbital robotics programs.

The tradeoff that governs the class was measured for the Shuttle arm program and has not changed. A harmonic drive gives a high single-stage ratio, negligible backlash, arc-second accuracy and negligible tooth wear, with transmission life set by surface fatigue of the bearing inner race. Against that, it needs additional torque to overcome static friction, exhibits substantial wind-up with a nonlinear torsional spring rate at low torque, tends to cog at low speed, and carries a wave generator whose inertia may be too high for a fast servo. Efficiency falls from the usual 80 to 90 percent with a wet lubricant to as low as 50 percent with none [1]. The same program also found the internal ring gear output to be twice as stiff as an external ring and pinion with better static and dynamic friction, and recommends keeping output shaft power below half the rated motor peak power.

Gearsets and actuators selected and flown

Part	Manufacturer	Used by	Source
CSF 14-80LW	Harmonic Drive	ipex (wheel actuators)	[2][3]
CSF 20-160LW	Harmonic Drive	ipex (bucket drum and arm actuators)	[2]
SHG 32-160	Harmonic Drive	rassor (shoulder actuator)	[4]
80:1 harmonic drive transmission	United Shoe Machinery	apollo-lrv	[5]
Carousel drive, 2500:1	Not named	curiosity (SAM)	[6]
Elevator drive, leadscrew	Not named	curiosity (SAM)	[6]
FHAC-mini motor-gearbox unit	Harmonic Drive	justin (Rollin’ Justin mobile base)	[7]

Ratings and qualification results

• CSF 14-80LW, Harmonic Drive: a strain wave gearset, 80:1, cup type [2]. Ratings: Sizing load cases run to 7.45 N m at the gearbox output. Output bearings sized to fit over the flex cup, input bearings stacked close together and preloaded with a wave spring sized so the wave generator axial force could not overcome it [2][3]. Qualification: Ran 6,547,036 input revolutions in the 3-sigma accelerated life test against a mission requirement of 5.7 million, then failed. Root cause was the wave generator moving axially into the flex cup despite the wave spring, breaking the motor rotor bearing springs. The future work identified is a change to the grease application on the harmonic drive to improve lubrication between the flex cup and the wave generator band [2].
• CSF 20-160LW, Harmonic Drive: a strain wave gearset, 160:1, cup type [2]. Ratings: Driven by a ThinGap LSI 75-30 motor. The arm actuator adds a power-off brake to stop back-driving with power removed; the bucket drum actuator adds an internal shaft carrying torque through to a second drum [2]. Qualification: Same four-stage KSC qualification campaign as the wheel actuator, run at 2-sigma rather than 3-sigma load. Results not published [2].
• SHG 32-160, Harmonic Drive: a strain wave gearset, 161:1 [4]. Ratings: Direct-driven by a Parker Bayside K089050 frameless motor, with a THK cross roller bearing carrying radial, axial and moment loads in one element [4]. Qualification: Selected against the requirement to lift and lower the arm with a bucket drum full of regolith at the required output speed. The paper reports the sizing basis and the fabrication lessons, not a test result [4].
• 80:1 harmonic drive transmission, United Shoe Machinery: a strain wave gearset in a hermetically sealed wheel drive unit [5]. Ratings: One per wheel. Motor and wave generator sealed together in dry nitrogen at about 5.17 N/cm2 and lubricated with Krytox 143AZ oil [5]. Qualification: Flown on Apollo 15, 16 and 17. Selected as baseline over an open-to-vacuum 80:1 planetary spur alternate from General Electric, which the contract required be carried in parallel until the sealed unit was shown to work under simulated lunar conditions [5].
• Carousel drive, 2500:1, vendor not named, built for the SAM Sample Manipulation System: a 25:1 planetary gearbox and a 100:1 harmonic drive behind a Hall-commutated brushless DC motor [6]. Ratings: Overall 2500:1 with 30,000 encoder counts per revolution. Positions both the sample carousel and the sample cup elevator from one actuator, the carousel being locked to ground by a spring-energized latch while the elevator is moved. Qualification: The whole Sample Manipulation System draws under 5 W continuous and 10 W peak and recovers from unexpected power loss through its feedback devices and control architecture [6]. In flight, Teflon and molybdenum disulfide lubricant particles from the sample chain were found to pass the 150 micron sieve [6].
• Elevator drive, leadscrew, vendor not named, built for the SAM Sample Manipulation System: a compliant linear actuator with a resettable launch lock, driven through an 80:1 planetary gearbox, a 2.4:1 spur pair and 5/16-12 leadscrews [6]. Ratings: Raises and lowers sample cups, supplies the force to seal cups into the pyrolysis ovens, and provides the launch restraint for the elevator mechanism. Qualification: Flown since August 2012 as part of the mechanism above [6].
• FHAC-mini motor-gearbox unit, Harmonic Drive: an integrated brushless DC motor with 100:1 harmonic drive [7]. Ratings: Used as the steering drive of the wheeled base [7]. Qualification: Ground system. The paper documents the selection and the 100:1 reduction; no space qualification is claimed [7].

Where strain wave gearsets fail

A wave spring sized against the wave generator axial force did not hold it. The IPEx wheel actuator input bearings were stacked close together and preloaded with a wave spring whose force was calculated so that the axial force generated by the Harmonic Drive CSF 14-80LW wave generator could not overcome it, keeping the motor rotor and wave generator axially positioned [3]. In the 3-sigma accelerated life test the wave generator moved axially into the flex cup regardless, breaking the springs on the motor rotor bearings and repeatedly tripping the motor current limit [2]. The actuator still reached 6,547,036 input revolutions against a mission requirement of 5,700,000 before the test was stopped, over multiple inspections and disassemblies. The future work identified is not a stiffer spring but a change to the grease application on the harmonic drive, to improve lubrication between the flex cup and the wave generator band.

Both harmonic drive failure regions recorded above are at the wave generator, not the gear teeth. The Shuttle arm characterization independently places transmission life at the surface fatigue life of the bearing inner race rather than at tooth wear, and records that wave generator inertia may be too high for a fast servo [1]. Sizing a strain wave gearset against output torque alone addresses neither the IPEx failure nor the life limit the earlier program measured.

Gearsets, springs and actuators with a published test result

Seven further mechanisms carry a measured result and no flight heritage on a planetary robot. Three of them are cases where the article measured far from what the design calculation predicted [8][11].

Part	Manufacturer	Used by	Source
Size 40 S-tooth harmonic drive	Harmonic Drive	Mars rover actuator study	[8]
EH-3525 paraffin actuator	Sierra Nevada	dsoc docking mechanism	[9]
DM105 docking mechanism	Not named	dsoc engineering model	[9]
Cu43Zr43Al7Be7 spur gears	Not named	KSC dynamometer article	[10]
L-PBF Ti-6Al-4V spiral spring	Not named	JPL torsion stand	[11]
Suspension actuator, 360 mm stroke	Not named	viper	[12]
Percussion mechanism	Not named	perseverance (coring drill)	[13]

Ratings and qualification results

• Size 40 S-tooth harmonic drive, Harmonic Drive [8]. An alternative tooth form in the standard size 40 envelope. Ratings: 3.0 kg unlightened, the same as the standard size 40 and half the size 50 [8]. Qualification: ratchet torque limit measured on a dynamometer at 1580 N m, against 745 N m for the standard size 40 tooth form and 1410 N m for a standard size 50, with a momentary peak limit of 1370 N m [8]. The tooth form more than doubles the ratchet limit at constant mass, which is a gain the standard catalog reaches only by going up a frame size and doubling the mass.
• EH-3525 paraffin actuator, Sierra Nevada. High-output thermal actuator in the Deep Space Optical Communications docking mechanism. Ratings: paraffin melting range 86 to 89 C, a vendor property not measured in the work that uses it [9]. Qualification: the vendor’s stated temperature limit before seal rupture is 130 C, corresponding to a 70 lbf cut-off load at the actuator shaft, and was adopted as the design limit. A thermal model run peaked at 145 C with no damage found afterwards [9], so the limit was exceeded in analysis and not in test.
• DM105 docking mechanism, vendor not named, built for the same instrument. Ratings: one engineering unit, primary heater carrying 95 percent of the cycles at 27.4 to 33.6 V [9]. Qualification: 420 dock and undock cycles accumulated in thermal vacuum against a requirement of more than 100, of which 41 were 2 to 8 hour holds and the rest 25 minute holds, split 47 percent at room temperature and 26.5 percent each at the hot and cold protoflight cases [9]. Undocked hold time reached 335 hours against a 76 to 304 hour requirement, with no dithering in any hold and no reported performance degradation.
• Cu43Zr43Al7Be7 spur gears, vendor not named. Bulk metallic glass gearing. Ratings: compared against C300 maraging steel [10]. Qualification: 65 percent less unlubricated wear than the steel, reported as heritage from earlier alloy development with the conditions and test duration not given [10]. The comparison is by mass loss between materials of different density, which is the same metric defect that has inverted other wear rankings in this database.
• L-PBF Ti-6Al-4V spiral spring, vendor not named. Laser powder bed fusion printed spiral spring for a camera cover hinge, three designs at four specimens each on an EOS M290 with vendor stress relief only and an as-printed surface [11]. Ratings: design target 4.00 N mm/deg on design 17186799-1, against 4.60 to 4.69 calculated from the as-printed dimensions. Qualification: measured torsional stiffness averaged 0.606 N mm/deg on that design, 0.13 of the stiffness calculated from as-printed dimensions and about 0.155 of the design target, with quoted errors of 81 to 91 percent [11]. Stiffness was also direction dependent, higher winding inward than opening at p at or below 0.002, attributed to inter-winding contact and friction, and out-of-plane twisting of the cross-section was seen in most tests. Multi-jet fusion PA12 nylon springs of the same geometry measured within 9.8 to 12.9 percent of the spring-equation prediction, on one specimen per design read by hand with a protractor [11].
• Suspension actuator, 360 mm stroke, vendor not named, in the VIPER wheel module [12]. Ratings: a four-bar linkage giving quasilinear output from a rotary actuator over a 360 mm stroke [12]. Qualification: no life or environmental result is published. The torque transducer in the same module is on Sensors and IMUs.
• Percussion mechanism, vendor not named, in the Perseverance coring drill. Ratings: crank pin plain bearing designed to keep contact pressure below 3.45 MPa at the highest percussion output, reached by growing the shaft from 3.175 to 4.7625 mm and the 440C Grade 10 roller ball from 6.35 to 9.525 mm [13]. The shaft was nitrided to a 75 micrometer case at 65 to 68 HRC against a 58 HRC minimum ball, with 7.6 micrometer diametral clearance and 0.2 micrometer surface finishes, after early prototypes wore out roller bores and broke crank pins. Qualification: grease fill was cut from 10 to 15 percent of free volume to about 3 percent because at the higher fill the mechanism could not reliably start at the -70 C qualification minimum and the actuator stalled at its 10 A current limit, equal to 0.75 N m at the output shaft [13]. Braycote 600EF viscosity rises very nonlinearly between -50 and -70 C, so the cold start and not the running load set the fill.

Two of the seven entries above are a measured stiffness or torque that the design calculation did not predict, in opposite directions: the printed spring came out at an eighth of its calculated stiffness [11] and the S-tooth gearset at over twice the ratchet torque of the catalog form it replaces [8].