Interactive Decomposition of Large Assemblies

This thesis presents a selective disassembly planning system that is capable of constructing animated step-by-step instructions for solving a user-defined disassembly problem. In contrast to complete disassembly, the system focuses on detecting sequences that lead to the efficient removal of a limited number of specified parts. The assembled products for disassembly planning are represented by a list of polygon meshes that correspond to the individual parts contained in the assembly. The employed algorithms have been optimized to allow handling of complex products with a large number of parts and high geometric detail. The system is divided into two separate modules. The first module performs detailed analysis of the product and extracts relational information that can then be assessed in an interactive planning application that constitutes the second module. By utilizing the parallel computing capabilities of modern graphics processing units, we achieve high performance during assembly analysis. A tolerance mechanism has been incorporated into the system to account for imprecisions and artefacts in the available mesh data. Computed disassembly sequences consist of translating motions that result in the exposure of all specified parts for removal. A graphical user interface provides means for visualization and extensive editing of disassembly sequences. By restricting the space of tested motions to a discrete set of translations, the system can provide immediate visual feedback to all changes made by the user. During animation of the step-by-step instructions, visual cues are employed to highlight important aspects of the disassembly procedure. The assemblies that have been examined in our system contain up to 512 parts and 700,000 triangle primitives.