Robotics Case Studies
Complex Assemblies: Rapid Automated Generation of Correct Assembly Sequences
by Terri Calton
Sandia National Laboratories
Manufacturing companies throughout the world are rapidly changing in order to survive in today's highly competitive market environments. Some examples of coping with changing environments are manufacturing globalization, automated and intelligent manufacturing, virtual manufacturing, and agile manufacturing. The objective of the movement in manufacturing is to improve flexibility, reliability, and productivity, and to achieve competition-based technology development.
Accordingly, the main focus of Sandia's R&D program is to provide intelligent software tools that automate many of the manufacturing processes that have traditionally been known to be the most costly, the most time-consuming, and the most error-prone. Some of these include part-level assembly planning, fixture planning, grasp planning, motion planning, and tools planning. Our overall strategy to reduce these costs is to push the breadth of application and depth of analysis and to find the appropriate balance between human and machine planning. Our ultimate goal is to improve profitability of operations by developing smart software.
Sandia National Laboratories' Intelligent Systems & Robotics Center (ISRC) developed Archimedes, a planning and visualization software tool used to generate, optimize, verify, and examine sequences of assembly. Archimedes allows product and process engineers to define assembly process constraints, automatically determine alternative assembly sequences satisfying those constraints, optimize the assembly process according to a user-specified quality metric, and visualize and communicate the results quickly and graphically.
Standard industrial procedure is for an engineer to design a device using CAD (computer-aided design). The engineer would then give the design to manufacturing, who would build the piece parts, join them to make subassemblies, and combine these to make higher-level assemblies. Sometimes, parts and assembly sequences were designed that could be put together in virtual reality but might not actually work in real life. Also, at some point in the process, a part that had to be somewhere inside the assembly could have no way to get in, forcing the manufacturing engineer to backtrack and find the point in the procedure to get that part in. This system of operation is expensive and time consuming.
Archimedes generates sequences of assembly and checks them to ensure they are the most efficient. Once the engineer defines any assembly process constraints (e.g., cost, time limitations, etc.), Archimedes can systematically explore virtually all possible assembly sequences. It will automatically determine alternative assembly sequences and come up with the best process to satisfy the constraints. Results are communicated quickly and graphically. This graphic visualization enables the engineer to easily identify process requirements to add as sequence constraints. Because each assembly sequence is typically generated in minutes, the user can go through as many iterations of the constrain-plan-view-constrain cycle as needed to achieve an acceptable plan.
Archimedes was used to plan the assembly of the AIM9X missile guidance system. The challenge in this project was that this assembly was much larger than any that had been previously analyzed by any automated assembly planner, including Archimedes. With 471 parts, 170 Mb of Pro/ENGINEER data, 55 Mb of ACIS data, and 800,000 facets, this assembly was at least 10 times as large as any assembly previously planned automatically.
AIM9X missile guidance system Target array assembly
Early in 1999, the Archimedes team set yet another record - the automated assembly planning of a target array assembly used as a missile interceptor. This assembly contained 1750 parts, 140 Mb of Pro/ENGINEER data, 18 Mb of ACIS data, and 1,800,000 facets. This feat surpassed the previously set record almost three-fold. In both exercises, an initial assembly plan was generated in under 30 minutes on an SGI Indigo R10000. Alternative plans were generated in under 3.5 minutes.
The next challenge for Sandia was to build a framework to combine automated assembly processes with Human Factors and Figure Models (HFFMs). All automated assembly planning software tools assume that the individual components fly into their assembled configuration and generate what appear to be perfectly valid operations, but in reality the operations cannot physically be carried out by a human. Is there room for a hand to actually set the part in its place? Can a hand holding a wrench fit around the assembly? Is there enough elbow room in which to maneuver? Human figure modeling algorithms may indicate that assembly operations are not feasible and consequently force design modifications; however, if they had the capability to quickly generate alternative assembly sequences, they might have identified a feasible solution.
To solve this problem, Sandia integrates HFFMs with automated assembly planning; this allows engineers to verify that assembly operations are possible and to see ways to make the designs even better. Archimedes has been integrated with Transom's commercially available human figure modeling software, Jack. A user can quickly analyze the assembly operations to be performed by a human. If it is unacceptable for human assembly, the user can quickly generate alternative operations using sound, proven geometric reasoning algorithms.
Terri L. Calton