Automation disassembly sequence generation based on visual recognition and rules in remanufacturing

Disassembly is a vital process for remanufacturing. Generating the feasible disassembly sequences for end-of-life products is an evitable step before executing the disassembly process. For most existing methods to generate the feasible disassembly sequences, CAD model of products should be provided. However, when the CAD model could not be provided, additional methods should be utilized to generate the feasible disassembly sequences. Especially for the situation that a part of this product is missing, the feasible disassembly sequence should be dynamically generated. In this paper, the disassembly rules are manually summarized. And, the feasible disassembly sequences are generated based on visual perception and rules. The method of deep learning is utilized to recognize the parts of end-of-life products. Finally, case studies under different scenarios based on double-axis diaphragm coupling are carried out to verify the proposed method.