Thermal Vacuum Testing

A thermal vacuum campaign serves three separate purposes that are often run in one chamber entry: bakeout and outgassing characterization, thermal balance for model correlation, and thermal cycling for workmanship and design verification. The verification argument for flight is carried by the correlated model, not by the test itself, because the test cannot reproduce the flight thermal environment [3].

Thermal balance and model correlation

Thermal balance establishes steady-state temperatures at bounding hot and cold cases so the thermal math model can be adjusted to match. On the Magnetospheric Multiscale observatories, three balance cases were run, hot operational, cold operational and cold survival, in a chamber held below 1.3 x 10^-4 Pa with the shroud at -180 C. Equilibrium was declared when the control sensors sat within 2 C of the plateau goal and the energy balance term mCpdT/dt fell to 2 to 5 percent of the total thermal control subsystem energy, the criterion GEVS sets [1]. Instrumentation was over 400 one-wire sensors, 200 thermocouples and 125 flight thermistors. Balance testing on the first observatory took approximately fifteen days, with the hot balance criterion reached about 24 hours after the configuration was established [1].

Correlation targets and results on that campaign:

Correlation goal	Criterion
Mean and standard deviation of model minus test	Within 3 C
Individual temperature error	Within 5 C
Heater power	Within 5 percent of test data

Balance case	Model minus test achieved
Hot balance	Mean -1.3 C, standard deviation 1.9 C
Cold balance	Mean 0.8 C, standard deviation 1.7 C
Cold survival	Mean 0.2 C, standard deviation 1.4 C

Source: [1].

Correlation order matters: large system variables first (multi-layer insulation effective emittance, conductive interfaces at structural rings) to bring the vehicle average into line, then individual box temperatures through component MLI emittance and interface conductance. Transient eclipse correlation additionally requires reconciling modeled thermal mass against measured mass and revisiting specific heat assumptions, since aluminum 6061 is typically assumed by default [1]. Test margin was demonstrated by extending the four-hour survival eclipse to five hours and the two-hour operational eclipse to three [1].

Correlation is the schedule driver. A large model with hundreds of sensors can take several months to correlate. The VIPER thermal team decomposed the integrated thermal model into faster-running breakout models that 14 analysts correlated in parallel, completing the integrated model in 22 weeks and reaching an overall RMS error of 7.4 C [3].

Bakeout and contamination

Bakeout accelerates water vapor outgassing and drives off species from polymeric and epoxy materials on both the article and the ground support equipment. On the MMS campaign each observatory sat inside a 40-cryopanel thermal enclosure, and the sequence was to pump down slowly, with several deliberate represses to sweep out water vapor, before flooding the scavenger plate, activating the quartz crystal microbalances and the residual gas analyzer, and stair-stepping the enclosure and shrouds up to the bakeout setting of 50 C so the observatory always ran warmer than its surroundings [2]. The microbalances were held at -20 C and the pre-test goal was a delta frequency decay to roughly 200 Hz/hr, an order of magnitude below the requirement [2]. Bakeout continued until those readings met the goal with margin and pressure, driven mainly by water vapor, had dropped far enough. The whole profile from pump down to vent back ran about 25 days [2].

Instrument constraints set the schedule around the bakeout. The instrument suites use microchannel plates and solid state detectors, which are sensitive to molecular contamination and moisture, so high voltage operation was gated on the enclosure interior holding below 1 x 10^-6 torr [2]. Reaching that pressure with a fully dressed observatory and its ground support equipment inside is what the 50 C bakeout was for: the empty chamber alone reached the low 10^-7 torr range on its cryopumps [2].

Accelerated cycling against the flight cycle count

Cycling for workmanship and design verification is run accelerated, and the acceleration is the part of the test that has to be justified against the mission. The Mars 2020 Package Qualification and Verification test defines one thermal cycle life as 350 cycles between 0 and 80 C plus 150 cycles between -40 and 40 C, and solder joints on a flight-like NVMCAM flash memory slice cracked at 500 cycles of 80 C delta T, before one thermal cycle life was complete [9]. No sample size, lot or date code is reported for that slice [9].

The Curiosity Rover Compute Element carries the same board design, so its flight thermal history can be counted on the same axis. Rainflow counting of about 3 million platinum resistance thermometer samples, converted between delta T ranges by Coffin-Manson with a solder fatigue exponent of 2 and with ground and cruise cycles excluded, gives 144.21 equivalent 80 C delta T cycles on the NVMCAM-A board and 429.42 on NVMCAM-B over surface operations [9]. The dominant diurnal cycle at the board on the Martian surface is about 15 C delta T with the cycle population concentrated between 10 and 20 C, so the count is built almost entirely from cycles much smaller than the test cycle and rests on the Coffin-Manson exponent to convert them.

NVMCAM-B reported temperature only intermittently, and its series was reconstructed as the mounting interface temperature plus a fitted multiple of the difference between the RAD750 and interface sensors, with 2.38 percent average and 0.67 C RMS error against the measured points that do exist [9]. That error is largest near 20 C and smaller at the high temperatures that drive fatigue.

Put side by side, the two numbers rule out the mechanism the test was run to bound. RCE-A failed to mount its flash on sol 200 at 144.21 equivalent cycles, against the 500 that cracked joints on the ground, so thermal fatigue does not account for that failure [9]. The comparison does not close either way, because no failure was reproduced on the ground and no cracked joint from a flight unit has been inspected [9].

What a thermal vacuum test does not reproduce

Gravity in two-phase devices. A loop heat pipe is orientation and gravity sensitive, so a chamber test at 1 g does not bound flight performance on its own [4]. The VIPER engineering design unit loop heat pipe was tested at the nominal orientation within the rover, vertical, and inclined so that the gravitational component acting on evaporator and condenser matched one sixth of the vertical case, with loop conductance used as the metric for the gravity effect [4].

Sky and ground view simultaneously. A correct lunar surface simulation needs a sky view near 4 K and a ground view anywhere from 40 to 400 K, set independently [5]. Lunar regolith at the equator averages 253 K and ranges 93 to 396 K; near the poles it averages 218 K and ranges 53 to 213 K; permanently shadowed crater floors sit near 43 K [5]. Equipment temperature is set by the radiative and internal heat loads and by the surface ratio of solar absorptance to infrared emittance, not by the regolith temperature. Polar temperatures fluctuate by more than 100 C between day and night and can fall to -246 C in permanent shadow [4].

Vacuum level with a dusty article inside. This is the sharpest limit. The lunar surface pressure is at or below 1e-9 torr, while a chamber holding a large mass of particulate cannot reach it because the dust itself outgasses [5]. The Apollo Lunar Roving Vehicle wheel tests were run in a chamber containing about 3000 kg of particulate at a pressure no lower than about 1 torr; a limited dust sample has been taken to 1e-8 torr; the practical target for a dusty chamber is on the order of 1e-7 torr [5].

The consequence is physical, not procedural. Adsorbed gas molecules on the dust and on the test surface disrupt the interaction between them. The fractional surface coverage of adsorbed gas that would have to be reached to avoid this is not a single number: it is a function of the chemistry of the dust surface, the chemistry of the test surface, the gaseous species and the temperature. Below that coverage the particles behave as clean bodies, and the difference from atmospheric behavior is called the clean body effect [5]. Cohesion-dominated properties reach clean body behavior early: during the 1971 LRV wheel tests, LSS-4 simulant showed lunar-like cohesion at any pressure below 5 torr, with clumps thrown up by the wheel holding together until they struck the chamber wall. Adhesion to sensitive surfaces requires a much better vacuum, and how much better depends on the surface pair.

Simulant surface chemistry. A simulant loaded into a chamber carries adsorbed terrestrial water vapor and oxygen, and its transition metals, iron above all, are oxidized where lunar regolith is reduced. In situ processing is therefore part of the test: a controlled pump-down to 1e-6 torr slow enough to prevent dust explosions and fluidized bed effects, resistance heating to about 200 C, above the roughly 125 C lunar maximum, held for several hours until the mass stabilizes, then repressurization to about 1e-4 torr for a radio frequency air plasma treatment, then pumping back down [5].

Dust ingress into the vacuum plant. The Lunar Dust Adhesion Bell Jar at Glenn is pumped from near the top with the vent opening above the dust reservoir so that ballistically falling dust cannot reach the pump, through a horizontal vent section containing five baffles with alternating top and bottom windows, glass sampling tubes between the baffles to monitor transported dust, and a final downward bend into an antechamber whose pump vent is offset from its inlet. An oil diffusion pump with a liquid nitrogen cold cap was chosen because it tolerates dust ingress; a quadrupole residual gas analyzer in the chamber detects backstreamed pump oil at a partial pressure near 1e-11 torr [5].

The pressure a mechanism life test is run at. Spacecraft mechanism life testing is specified in a chamber better than 1.3 x 10^-4 Pa, equivalently 1e-6 torr, held for more than twice the total expected operating cycles at the mission temperature extremes [10]. That is the pressure at which flight lubricant and bearing heritage is generated, and it is six or more orders of magnitude above a lunar surface or an interplanetary cruise, so the chamber is still supplying the residual gas that re-forms the oxide a dry sliding contact depends on. The reproducibility floor on the resulting numbers is a 30 percent coefficient of variation between laboratories running the same standardized wear test under ASTM G83-90 [10].

Polar volatiles. Testing at polar conditions requires the simulant to be doped with water to a representative fraction and then chilled cryogenically under vacuum, which desiccates it. A 1 m tall, 28 cm diameter simulant bin with liquid nitrogen cooling loops at 100 K reached an average bed temperature of 140 K at vacuum, and pre- and post-test sampling was used to quantify the desiccation because no good in-situ method for measuring it under vacuum exists [7].

When the chamber gave the wrong answer

The limits above are arguments from physics. Two cases on flight vehicles are measured discrepancies, where the test was passed and the flight result differed.

A missing flux term can invert the conclusion. The Mars 2020 system thermal vacuum test in the JPL 25-Foot Space Simulator showed that one cruise thruster catalyst bed heater could remain on continuously from launch [11]. In flight it could not. The chamber simulated no off-Sun angle flux, and that missing term was enough to reverse the result; flight performance matched the corrected model rather than the test [11]. The failure is not that the chamber was imprecize. A heat load absent from the chamber was absent from the conclusion drawn from it. The same test report states outright that its temperatures are not a direct indication of flight performance, and that flight predictions came from post-test model correlation rather than from measured values [11].

The catalyst bed warmup times that campaign did measure carry their own configuration. On one heater string, the cruise thruster catalyst bed reached its 70 C hot fire temperature in 21 to 25 minutes, the descent reaction control system bed reached 25 C in 27 minutes in the cold balance case, and the Mars landing engine bed took 4 hours, which is why flight energizes both of its heaters just over a day before touchdown [11]. The 70 C thruster allowable is a requirement, raised from the 50 C of the MSL thrusters it replaced, not a demonstrated survival limit [11].

The test article’s own cabling is part of the test. Perseverance’s thermal vacuum campaign was run with ground support equipment cables, which lose 4 percent of voltage and 8 percent of power, and the correlated thermal model inherited that figure [12]. The flight flex cables lose 9 percent of voltage and 18 percent of power. The model consequently predicted preheat ramp rates with errors above 5 C in flight; after the cable loss was corrected, model and telemetry agreed within 5 C [12]. A correlated model is correlated to the test configuration, and every difference between that configuration and the flight one propagates into flight predictions silently.

Two build-to-print vehicles do not land on the same temperatures either. Comparing identical propellant line thermal zones between the MSL and Mars 2020 descent stages in the same chamber gave minimum-to-maximum range differences of 2 to 13 C, averaging 6 C, attributed to multi-layer insulation, installation and heater resistance workmanship [11].

What the flight vehicle then reported

Perseverance’s first 700 sols give the other side of the comparison, from flight telemetry rather than a chamber [12]:

Quantity	Value
RAMP avionics panel margin, hot side	11 C
RAMP avionics panel margin, cold side	49 C
RAMP diurnal range	30 C
RAMP rise during MOXIE operation	at least 5 C
External component non-operational allowable	-128 C

Source: [12]. Margins are to design allowables, not to demonstrated failure limits, and the values are per-zone examples from selected sols rather than a full statistical characterization.

The RAMP diurnal range is driven mostly by avionics dissipation rather than by the environment. The MOXIE rise was large enough that the thermal team constrained mid-sol MOXIE operations [12]. The -128 C external allowable is a qualification limit set to match the environmental minimum, so those components need no survival heating; cameras and actuators typically carry a -55 C minimum operational limit set by their internal electronics [12].

The Apollo brushing case

Nine dust-caused hazards were identified from Apollo experience: obscured vision, clogged equipment, coated surfaces, loss of surface traction, inhaled dust, degraded radiators, fooled instruments, failed seals, and abrasion [5]. Radiator degradation produced a documented failure of ground-test fidelity.

A pre-mission study using soil returned from Apollo 12 concluded that a nylon bristle brush was an effective way to remove lunar soil and dust. Brushing was used on the Lunar Roving Vehicle radiators during Apollo 15, 16 and 17 and was found to be almost wholly ineffective; the brush did not remove the finest particles and radiator thermal performance degraded despite the brushing [5][6]. The discrepancy is attributed to the ground test not reproducing the lunar surface condition of the dust and the surface.

Re-testing the same technology in the Lunar Dust Adhesion Bell Jar under simulated lunar conditions was done as a four-stage campaign: ambient screening of strip brushes in nylon, PTFE and Thunderon bristles, then the best performers in the chamber, then a wider set of brush geometries at ambient, then the best of those in the chamber. Two brushes, a Zephyr fiberglass fingerprint brush and an Escoda nylon fan brush, removed over 90 percent of the dust from AZ93 and aluminized FEP, recovering 80 percent or more of the original solar absorptance to emittance ratio after 20 strokes and 90 percent or more after 200 strokes [5][6]. Absolute brush performance depends on which simulant is used, so the transferable result is the relative ranking rather than the percentages [5][6].

Dust tolerance beyond thermal surfaces

There is no standardized, widely accessible method for evaluating a material or a mechanism in a facsimile of the lunar surface environment, which is itself a brake on dust-tolerant technology. Testing that puts abrasive regolith, vacuum, temperature and radiation on an article at the same time is what is missing; the SLIDE testbed at Langley addresses part of it by combining controlled lunar simulant deposition with a chamber pumped to the 10^-6 torr level, in an 86 by 46 by 51 cm working volume [8].