By combining the design optimization capabilities of the ANSYS simulation program with a proprietary loads analysis routine, Johnson Engineering Corp. (Webster, TX) created a better robot arm for astronaut training. The new arm, called WRMS (Weightless Environment Training Facility Remote Manipulator System), weighs 300 pounds less than its predecessor. Yet it can lift 200 pounds compared to the fifty-pound limit of the earlier device.

 Working with engineers at Engineering Cybernetics Inc., a San Antonio, Texas-based consulting firm, Johnson Engineering used a variety of simulation methods in the redesign of the robot arm. Some were traditional in nature, such as the nonlinear buckling and contact analyses performed with ANSYS design analysis and optimization software from ANSYS Inc. (Canonsburg, PA). Combining an in-house loads program with the design optimization capability of ANSYS, however, was a bit of creative engineering that went a long way toward reducing weight and boosting the lifting capacity of the arm.

Replica of the Real Thing

 Although similar in operation to the Space Shuttle arm, the design of the WRMS in most respects is completely different. It would be impossible to duplicate the design of the actual Space Shuttle arm on earth since it was designed to operate in zero gravity. In fact, the Space Shuttle arm cannot even pick itself up in 1G.

 "Our challenge was to simulate the operation of the Space Shuttle arm during flight, but the two arms are actually very different," explains David McMahon, a Johnson Engineering project engineer who works at the Johnson Space Center. McMahon was in charge of the structural design of the arm. "The WRMS has the same volume and controls like the real thing, but it is run with hydraulic power instead of electric power," he adds.

 One of the complications facing Johnson Engineering in this project was the many different positions the device could assume. To evaluate lifting capacity, it was necessary to calculate torque at thousands of different positions of the wrist, elbow, and shoulder joints. To complicate matters further, at certain positions some portions of the arm rise above the surface of the water.

 Johnson Engineering wrote a program to handle these calculations, including the changing buoyancy conditions as the arm position changed. For each input (any position of wrist, elbow, and shoulder joint and one of two different loads), the program computed forces and torques on the joints.

 To automate the investigation of the many possible arm positions on lift capacity, engineers on the project combined the in-house loads program with ANSYS. "This was an unusual way to use the optimization capabilities of ANSYS," admits Doug Scheiding, engineering manager at Engineering Cybernetics. "We were using the pre and postprocessing capabilities of ANSYS with the loads program as the solver instead of running an FEA simulation."

 Typically with automatic design optimization, a user specifies an engineering goal or objective (such as reduced weight or a target load capacity), variables (thickness of members, material type, geometry), and design constraints such as permissible stress levels. The software runs through all possible combinations of variables to find the design that comes closest to the user's requirements. During the optimization, the software solves a series of finite-element analyses to evaluate the performance of different design possibilities.

 For the robot arm optimization, however, ANSYS was not performing a finite-element analysis with each design iteration. It was using the results from Johnson Engineering's loads program as though they were analysis results. "Our program gave ANSYS the forces and torques, and it used that information to determine the maximums at each joint," says McMahon. "This really helped us reduce weight because it made it possible to evaluate thousands of positioning possibilities."

 In addition, the engineers used ANSYS to perform a nonlinear buckling analysis of the arm's two booms. Johnson Engineering had specified titanium for the booms because of its excellent strength-to-weight ratio and its corrosion resistance in the chlorinated training tank. The analysis was needed to optimize the wall thickness of these members for the given operating loads.

 Scheiding and colleague, Philip Poll, of Engineering Cybernetics performed these analyses, working from Johnson Engineering's Pro/ENGINEER models of the booms. After converting the CAD data to ANSYS, they ran a large deflection buckling analysis on the booms using the maximum load of 200 pounds. These nonlinear analyses were run on a four-processor SGI Challenge computer, taking advantage of the parallel processing version of ANSYS, reducing solution time to less than 30 minutes for each optimization study.

 An optimum wall thickness of 0.080 inch was established from the results. Scheiding also used the nonlinear buckling analysis to simulate a large dent in the boom, to evaluate the possibility of the arm being damaged during use.

 All analyses were done while the robot arm design was in progress. Several parts of the arm, most notably the three pieces comprising the elbow, underwent major changes when analysis showed the original design would not meet stress and deflection requirements. The tight connection between ANSYS and Pro/ENGINEER, along with the ANSYS Parametric Design Language, made it easy to evaluate additional design options, says Scheiding.