You are on page 1of 17

Proceedings of the X on Mechanical Engineering December 6, 2012, Pittsburgh, PA, USA

STABILITY AND ENERGY ABSORPTION EVALUATION OF A LUNAR LANDER BY 1/6 SCALED DROP TESTING AND FULL-SCALE SIMULATION

Ahmet Sahinoz Mechanical Engineering Carnegie Mellon University Pittsburgh, PA 15213 asahinoz@andrew.cmu.edu

Mikio David Mechanical Engineering Carnegie Mellon University Pittsburgh, PA 15213 mikiod@andrew.cmu.edu

ABSTRACT Stable soft landing is crucial for planetary missions. Shock and tip over are risks that are mitigated by spacecraft landing gear. This research evaluates stability and energy considerations for landing gear. 1/6 scaled prototype and full-scale simulation of an 850kg lander were used to determine landing characteristics for various touchdown scenarios. A 1/6 scale lander model is designed and prototyped using appropriate scaling factors to simulate full scale landing on the moon. Drop tests are performed in comparison with simulations to investigate landing loads, honeycomb stroke and tip over stability on different slopes and surface materials. Remarkable correlation is achieved. The system provides a safe landing under the worst case scenarios that are based on the current requirements and previous lunar missions. Scaled prototype stably landed on a lunar simulant at 30deg downhill slope with 4m/s vertical and 1m/s horizontal velocity. Lunar simulant absorbed 50% of total energy on average. 1 INTRODUCTION Planetary landers require compliant legs to touchdown undamaged in a stable position, ready for operation. The proper understanding of the mission requirements, reduced gravity forces, lander mass properties, worst case touchdown scenarios and the stowing space limitations of the launch vehicle are of great importance in order to design an optimal landing system. Challenges are to design for uncertain landing conditions and to perform tests on Earth to simulate lunar gravity. Landing gear must cope with the expected mass, velocity and orientation of the lander at touchdown, in the expected range of terrains, and doing so with minimal mass and a margin of safety. Uncertainties include the mechanical properties of regolith, slope of the surface and rock distribution.

Fig 1: Scaled model with Astrobotic Griffin lander (mock up legs) and Red rover During final descent, the vertical velocity of the lander will be reduced nearly to zero as a result of the deceleration provided by the main thruster. Hazard detection identifies obstacles larger than a threshold value and selects a landing site. A horizontal velocity component may be present due to detection errors or hazard avoidance maneuvers. The main engine cuts off at a predetermined altitude in the order of a few meters to prevent instability due to surface effects. The touchdown occurs following a short free fall phase. Resulting kinetic energy has to be dissipated over a finite distance while providing sufficient clearance and a stable landing [4]. In this paper, landing conditions and spacecraft landing system are described first; and then, landing loads, energy absorption and stability are investigated through full scale simulation and scaled drop experiment results.

2 BACKGROUND Energy absorption approaches for legged lunar landers include spring-damper systems and crushable materials. Apollo landing module utilized a cantilever leg design with aluminum honeycomb cartridges in all three struts of each leg. The main strut has a compressive stroke only, whereas lower struts incorporate a tensile stroke because they can experience both types of loading [3, 5].

10cm with 1 to 4m/s vertical velocity with about 300kg landed mass. Soil bearing strength increases with depth, starting from 1.8N/cm2 at 1-2cm and reaching 5.5N/cm2 about 5cm. Friction coefficient is reported in the range of 0.3-0.7, representing an equivalent value to be used in rigid simulations [1, 2]. Lunar simulants are available with different particle sizes and compositions to investigate the effects of regolith on landing dynamics [9].

Fig 4: Sub-scale drop testing illustration Stability evaluation methods include simulations, sub-scale testing and full scale experiments. Simulations provide preliminary insight on landing characteristics. Sub-scale testing can validate energy absorption and stability when appropriate scaling factors are used to achieve dynamic similarity. Full scale testing is necessary for final assessment and verification of structural design. 3 PERFORMANCE REQUIREMENTS The problem is defined by the specific requirements for the mission, including lander properties and lunar conditions [6]. Landed mass of the spacecraft is in the range of 650-850kg depending on the fuel leftover and payloads. The vertical acceleration limit of the structure is defined to be 10gs for Earth gravity (9.8m/s2). No strict limitation is put on the horizontal component. Dynamic envelope of the launch vehicle (Falcon 9) is 4.6 meters in diameter that the legs must be contained within. The hazard avoidance system can detect rocks larger than 25cm with some error, defining the required clearance. Therefore, the bottom of the lander should stand at least 25cm higher than the ground after the landing. Slopes that are larger than 10deg can be detected, thus, the lander may have 10deg error with respect to the ground during touchdown. Maximum expected slope is 20deg. Angular velocity is assumed to be negligible. Nominal landing velocities are 2m/s vertical with zero horizontal component; to be conservative, the worst case scenarios are selected as 4m/s vertical with 1m/s horizontal velocity, considering Surveyor landings as a reference. Lunar gravity is 1.63m/s2, 1/6 of Earth. Temperature has a wide range from -100 to 120oC depending on the day time and the region. The effect of temperature is neglected in this study. 2

Fig 2: Apollo landing gear and main strut design Surveyor landers used an inverted tripod leg design with shock absorbers. The legs rotate about the hinge axis, compressing the shock absorber which dissipates most of the spacecrafts kinetic energy. Then the legs re-extend, returning to their pre-landing configuration. Aluminum honeycomb blocks under the body are utilized as a secondary absorption system if there is excessive kinetic energy. The footpads made of aluminum honeycomb function as a load limiter in case the pads land on a rock. [8]

Fig 3: Surveyor landing gear design (one of three legs) Mechanical properties of lunar soil are important factors for landing system design. Bearing strength determines the requirements on footpad size, given the expected force and allowed penetration distance. The friction force is also determined by these properties, footpad profile and landing velocities. Surveyor had a 30cm footpad and penetrated 2 to

4 SPACECRAFT LANDING SYSTEM The primary structure of the lander consists of a deck, upper cone, lower cone and bulkheads illustrated in gold. A rover sits on top of the upper cone; fuel tanks are located within four holes on the deck; weight of the tanks is transferred to the bottom ring through bulkheads. Rectangular plates near the bottom ring between bulkheads are the hinge connection points for the lower struts of legs.

Kinematic analysis showed that for a vertical landing on a flat rigid surface with friction coefficient of 0.3 (min expected), the maximum total energy can be dissipated over 10cm vertical displacement of the center of mass with 12kN honeycomb in each leg. The honeycomb crush corresponding to this displacement is 8cm. This is related to the angle of the main strut with respect to the vertical axis; higher the angle, lower the stroke value. Although 8cm crush length is sufficient for an ideal landing, 25cm honeycomb (crush length 17.5cm, 70%) is used to provide sufficient stroke for the worst case of landing on a rock.

Fig 7: Main strut (transparent) with honeycomb insert The wider the stance compared to the center of mass height, the more stable the lander is in the presence of horizontal velocity, orientation errors and uneven terrain. Footpad placement is done such that the static stability angle is 42deg, 10deg less than Apollo, providing a lower profile. Base width is 3.6m, equal to the diagonal across the deck. Top of the footpad is located 40cm below the bottom ring, providing 30cm clearance after a maximum displacement of 10cm for crush. The height of the footpad is not added into this clearance because it is expected to penetrate into the lunar soil.

Fig 5: Lander structure with telescoping legs Landing legs are inverted tripods with telescoping main struts. The motion during landing is a rotation about the hinge axis, shown in Fig 6, and the energy is absorbed by a honeycomb cartridge which provides constant force throughout the stroke. Assuming the lower struts provide high stiffness in the lateral plane, the main strut is under pure axial compression.

Fig 8: Illustration of lander parameters Footpad diameter is 45cm to support landing forces with penetration less than footpad height which is about 7.5cm, considering the bearing strength data of the lunar soil from Surveyor landings. The top section of the footpad is a dome, providing remarkable strength with the same mass compared to a flat surface. Honeycomb with 30psi crush strength is placed under the dome. The round profile with a thin sheet of aluminum bonded at the bottom minimizes penetration and friction.

Fig 6: Landing gear schematic Aluminum honeycomb with 360psi crush strength is used in 63.5mm diameter main strut of each leg, providing constant force of 12kN during the crush with 10% deviation [7]. 10g maximum acceleration requirement for a 650kg (lower limit) lander results in an approximate total maximum force of 64kN, 16kN for each leg. The energy that needs to be absorbed is equal to the kinetic energy of an 850kg (upper limit) lander with 4m/s velocity.

Fig 9: Footpad cross sections (w/o honeycomb) Honeycomb footpad stays intact if the landing is on regolith, but partially crushes if there are rocks on the surface, creating an even contact. It also deforms to soften the impact on the leg when excessive side loading is present, in case of hitting a rock while sliding on the ground. 5 METHODOLOGY 5.1 Simulation Honeycomb stroke, landing loads and tip over stability are investigated in a 2D motion simulation program. Two separate models are created for 2-2 and 1-2-1 scenarios. In simulation environment, lander structure and the ground are rigid. Leg angles are identical to the original design. Honeycomb is modeled as a constant force spring. Mass, center of mass and rotational inertia of the lander are defined. Parameters varied are lander velocities, slope and friction coefficient. Fig 11: Scaled lander prototype The inner diameter of the scaled main strut is equal to 12.7mm and the length is 35.6mm in which the crush material must be inserted. For this purpose, samples from 360psi and 690psi honeycomb are prepared and crushed with an Instron compression tester. 360psi samples crushed with an average constant force of 325N and the 690psi samples with 650N, both by 70% of their initial length before the force increased rapidly. Although the samples had small number of unit cells, they crushed very consistently. Test plots are located in Appendix B.

Fig 10: Simulation model (2-2 configuration) 5.2 Scaled Lander The need is to create a representative and a viable scaled test platform because a full scale drop test requires extensive use of resources. To achieve dynamic similarity to full scale landing on the moon, a 1/6 scale model is designed according to a set of scaling factors used in Apollo testing methods [10]. The reason behind the match is the similarity of acceleration/gravity ratios. Important parameters in design are leg geometry, total mass, center of mass height and rotational inertia. Scaling factors and parameters used are specified in Appendix A. Legs and the body are made of aluminum with bronze bushings and plastic footpads. A steel counterweight cylinder is placed on top and large diameter holes are drilled to the body to satisfy the desired mass property values within 20% error. After the complete assembly, total mass is measured to be 4kg which corresponds to 850kg in full scale. Other values are evaluated in software and are not physically measured. 4 Fig 12: Progressive crush of 360psi cartridge (3 unit cells) Struts have venting holes to minimize force increase due to air compression for the drop tests. 6 LANDING LOADS A set of experiments are conducted in CMUs Motion Capture Room, equipped with high speed infrared cameras distributed around the room that can track coordinates of the markers on the lander at 480fps with 0.1mm resolution. A total number of 10 landing scenarios with different configurations, velocities and pitch angles are tested. A four bar mechanism (Fig 4) with pitch adjustment is used with an electromagnet as a quick release to drop the lander. Real pitch angle values were different than predetermined values due to the imperfection of the drop rig. Thus, after experiment results

are generated, pitch angle and velocities are imitated in the simulations to have a meaningful comparison. The time frame of the landing is about 10ms for the scaled drop with 690psi honeycomb cartridges. 360psi honeycomb is chosen to extend the time period of crush to 20ms for the purpose of having more data points.

Fig 13: CMU Motion Capture Room (MOCAP) Friction coefficient on the ground is measured by dragging the lander with a spring scale as 0.5, and as 0.3 on plywood. These friction values are used in simulations. The floor is plastic and has compliance. Structural elasticity is also present, different from the simulation. Honeycomb cartridges are handmade and pre-crushed to a desired length in order to achieve a tight fit. They are inserted into the main struts and the strut is threaded onto the clevis. 9 markers are placed on the lander: 4 on each footpad, 4 at the corners of the body and 1 on the counterweight in an asymmetric position. The markers on the footpads are used to determine the instant of contact. Markers on the body are utilized to extract the height, velocity and acceleration of each corner, and to calculate the center of mass accelerations by taking the average. Although the center of mass is located 2cm above the surface of the body, this averaging method provides very close results and is viable. Marker on the counterweight is used to identify the orientation of the lander, and it shows the acceleration values around the rover. Pitch angle is calculated from two markers on the body using the height difference and the known distance between the footpads. Angular velocity and acceleration are derived from the pitch angle with a finite difference method. Presence of small roll and yaw angles are neglected in calculations. The time scale of the experiment is multiplied by 6, the linear accelerations and the angular velocity are divided by 6, and the angular acceleration is divided by 36 due to scaling factors in order to correlate with the full scale simulations. A full list of drop experiment landing conditions and comparison plots of five selected scenarios are located in Appendix A.

Results are plotted in both raw and lightly filtered form. Raw plots capture peak points more accurately but may contain false peaks and significant noise. Filtered plots provide a clean output, but decrease the amplitude at peak points. Output from the motion capture system is clean when the motions are significant compared to the size of the markers and the general direction of motion. Vertical position data is clean for each scenario, where horizontal data is clean only when significant horizontal motion is present, and angular data is clean when there is a notable pitching motion. Pitch angle plots of drop tests have oscillations that represent a bouncing and a rocking motion due to structural elasticity and ground compliance, where the simulation plots are perfectly rigid. For drops with an initial non-zero pitch angle, angular position, velocity and acceleration results are in good agreement. Vertical position, velocity and accelerations have good correlation, and the maximum acceleration observed is 7gs. Two acceleration peaks in Landing 3 and 5 (2-2 orientation) represent the consecutive impacts of two pairs of legs. Horizontal velocity tracked the simulation result in close proximity especially in Landing 4 and 5, but the noise in the measurement made horizontal data inaccurate for other cases. Maximum horizontal acceleration observed is 4gs. 7 ENERGY ABSORPTION Simulations showed that energy absorption and total force acting on the lander increase with friction coefficient of the ground. This variation is due to the kinematics of the legs, as the honeycomb crushes, legs move outward against a horizontal force that resists the sliding motion. This force also acts as an energy absorbing system and it is significant since the horizontal forces are between 30-80% of vertical forces depending on the friction coefficient. 12kN honeycomb force provides the desired energy absorption capacity at the minimum expected friction of 0.3, and maintains acceleration about 10g at the highest expected friction of 0.8. This addresses energy absorption characteristics for an ideal landing on a rigid surface. Landing on lunar regolith is a different problem which brings more variables into the field. Bearing strength is an important factor for penetration, which is a mode of energy absorption. Friction force depends on footpad profile and changes with penetration depth. Footpads penetrate the surface until the regolith compacts and reaches the honeycomb crush force, then absorption continues with crush until the energy is completely dissipated. Two lunar simulants are used to investigate the overall effect. Lander is first dropped on plywood at 3.7m/s to get a baseline for a hard surface, then, it is dropped on GRC-1 and JSC-1A in a soft and a compacted form. Average honeycomb crush at impact was 13mm on plywood, 9.5mm on compacted JSC-1A, 7mm on GRC-1, and 3.5mm on soft JSC-1A. 30-70% of the energy is absorbed by simulants. Maximum penetration is observed on soft JSC-1A which was about the footpad height. 5

Fig 14: Scaled lander on JSC-1A (soft) The most stroke intensive scenario is one leg landing on a large rock. This case is tested by placing a rock on GRC-1 with the maximum expected height, and dropping the scaled lander in a way that one footpad lands on it. The leg that hit the rock crushed 8mm and the second leg (diagonal) crushed 24mm where the other pair of legs barely made contact with simulant. The result is parallel with simulation.

Fig 16: Surface plot of marginal stability from simulation Internal region of the surface towards the left side in stable and external region towards the right side is unstable. A vertical landing at maximum velocity is always stable if the friction coefficient is below 0.5 and in all cases when slope is less than 20deg. At a friction coefficient of 0.8, front legs stick to the ground, and increasing the friction further practically makes no difference as far as tip over stability is concerned.

Fig 15: Rock drop test on GRC-1 The reason why the second leg absorbs the most impact is related to the ratio between honeycomb crush force and lander rotational inertia. Honeycomb force is high enough to convert the linear velocity of the lander into rotation with moderate crush; the lander has a small period of time where it rotates about the footpad, then the second leg makes contact with ground, absorbing the remaining energy. 8 STABILITY Tip over stability is first investigated by simulation. Surface plot of marginal stability is created for a vertical landing with zero pitch angle on a sloped surface for 2-2 orientation. Vertical velocity (0-4m/s), slope (0-45deg) and friction coefficient (0-1) are varied; data points are collected and plotted by linear interpolation. Color represents the velocity. Shape of the plotted surface is a tilted u-channel that is wider on top with increasing velocity, meaning it becomes more susceptible to tipping. 6

Fig 17: Landing stability boundaries for planar vertical landings on slopes, 2-2 spacecraft orientation The cross section plot of the 3D surface illustrates the two high velocity curves which have the same colors on both figures. Orange represents 3m/s and red corresponds to 4m/s. Simulation results are continuous lines with cubic interpolation, and experiment results are shown with markers. To have comparison with simulations, drop tests are performed on a stiff plywood surface covered with carpet, and high friction is achieved by using double sided tape on lander footpads. These tests produced similar results to simulations.

Fig 18: Marginally stable landing on a stiff surface To get more realistic results for landing on the moon, drop tests are conducted on a lunar simulant, JSC-1A. No lift-off is observed on trailing leg pair on 20deg and 30deg slope landings, and this indicated an equivalent friction coefficient of 0.5 in simulations.

was absorbed by GRC-1 compared to landing on plywood. JSC1A absorbed 30-70% of the energy depending on compactness. Honeycomb stroke is tested by landing one leg on a rock as this was the worst case. The first leg landed on the rock absorbed 25% of the energy and the opposite leg absorbed the remaining 75%. Two other legs did not make ground contact as expected. The second leg crushed 80% of its total stroke. Tip over stability is examined with drop tests on slopes with worst case velocities, first on stiff surfaces to relate to simulations, then on lunar simulants to generate realistic results. A surface plot of marginal stability is created with simulation. Drop tests on stiff surfaces with sticky footpads produced results that are similar. Then, JSC-1A is used to represent bearing strength and friction properties of lunar soil with high fidelity. Lander stably landed on a 30deg slope at maximum velocities on the simulant. The trailing legs did not lift-off and sliding was significant. From these observations, equivalent friction coefficient is estimated to be 0.5. Tip over stability is notable with the current design under the selected worst case conditions. Future work includes fabricating and testing full scale legs, and full scale drop testing to conclude the study. ACKNOWLEDGMENTS The authors thank Red Whittaker, Uriel Eisen, Justin Macey, Steve Huber, Kevin Peterson, William Pingitore, Jason Hallack, Eric Benson and Kevin Fulton for their support. REFERENCES [1] A. Ball, J. Garry, R. Lorenz and V. Kerzhanovich, 2007, Planetary Landers and Entry Probes, Part I, Chap. 7. [2] NASA, Surveyor Program Results, pp. 141-163 [3] Bryan, C., Strasburger, W., Lunar Module Structures Handout IM-5, NASA LSG 770-154-10, May 1969 [4] Buchwald, R., Witte, L., Schroder, S., Verification of Landing System Touchdown Dynamics, IAC-11.A.3.1.3, 2011 [5] Rogers, W.F., Apollo Experience Report - Lunar Module Landing Gear Subsystem, NASA TN D-6850, June 1972 [6] Astrobotic Technology, System Definition Review, 2010 [7] Plascore, 2012, Crushlite Lightweight Energy Absorption, http://www.plascore.com/pdf/Plascore_CrushLite.pdf [8] F. Sperling, J. Galba, A Treatise on Surveyor Landing Dynamics and an Evaluation of Pertinent Telemetry Data Returned by Surveyor I, NASA N67-34177, August 1967 [9] Simulant Working Group, Status of Lunar Regolith Simulants and Demand for Apollo Samples, December 2010 [10] Blanchard, U., Evaluation of a Full-Scale Lunar-Gravity Simulator by Comparison of Landing-Impact Tests of a FullScale and a1/6-Scale Model, NASA TN D-4474, June 1968

Fig 19: Touchdown instant on JSC-1A 30deg slope 9 CONCLUSION Landing characteristics are investigated in a 2D motion simulation program. Results provided valuable insight on the kinematics and dynamics of the problem. Simplicity of the method is novel, minimizing runtime. A scaled mass model of the lander is prototyped and drop tests are conducted in a motion capture room with various landing conditions to verify full scale simulation results. Pitching motions, center of mass velocities and accelerations were in remarkable agreement. The scaled prototype adequately reproduces landing dynamics and it is suitable for detailed studies. GRC-1 tests provided preliminary insight about the energy absorption characteristics on a soft surface. 50% of the energy

APPENDIX A: SCALING, DIMENSIONS AND TESTS Table 1: Scaling factors [2]

= Geometric model scale, = Gravitational ratio


Table 2: Lander parameters (1/6 scale model, corresponding full scale, real lander values)

Table 3: Landing scenarios of motion captured drop tests (pitch angle is affected by drop rig)

MOCAP Results Impact 1

Impact 2

Impact 3

Lift-off

Fig 20: Illustration of pitch angle plot for Landing 5

Fig 21: Landing 1 (Flat, = 0.5, Vx = 0, Vy = 3.3m/s)

10

Fig 22: Landing 2 (Flat, = 0.3, Vx = 0, Vy = 3.3m/s)

11

Fig 23: Landing 3 (Configuration: 2-2, Pitch = 7.5deg, = 0.5, Vx = 0, Vy = 3.25m/s)

12

Fig 24: Landing 4 (Configuration: 1-2-1, Pitch = 7.5deg, = 0.5, Vx = 0, Vy = 3.25m/s)

13

Fig 25: Landing 5 (Configuration: 2-2, Pitch = 11deg, = 0.5, Vx = 0.4, Vy = 3.25m/s)

14

APPENDIX B: HONEYCOMB BEHAVIOR

Fig 26: Representative aluminum honeycomb behavior [7]

Fig 27: Progressive crush of 360psi cartridge (3 unit cells)

15

In-strut Crush Test Results Initial length: 31.5mm, Force (N), Displacement (mm)

Fig 28: Instron compression test result of 360psi honeycomb sample

16

Fig 29: Instron compression test result of 690psi honeycomb sample

17

You might also like