Automated Decomposition of Complex Parts for Manufacturing With Advanced Joining Processes

Abstract

Abstract Advanced joining processes can be used to build-up complex parts from stock shapes, thereby reducing waste material. For high-cost metals, this can significantly reduce the manufacturing cost. Nevertheless, determining how to divide a complex part into subparts requires experience and currently takes hours for an engineer to evaluate alternative options. To tackle this issue, we present an artificial intelligence (AI) tree search to automatically decompose parts for advanced joining and generate minimum cost manufacturing plans. The AI makes use of a multi-fidelity optimization approach to balance exploration and exploitation. This approach is shown to provide good manufacturing feedback in less than 30 minutes and produce results that are competitive against experienced design engineers. Although the manufacturing plan models presented were developed specifically for linear and rotary friction welding, the primary algorithms are applicable to other advanced joining operations as well.