Disassembly Work Analysis Using Motion Capture and Machine Learning

In assembly manufacturing industry, Green Transformation (GX) from the linear economy to the circular one by reuse and recycling is required to prevent future depletion of natural resources and garbage problems. Thus, disassembly work plays an important role in this process, where an assembly product is separated into its components and/or subassemblies by non-destructive means. However, manual disassembly tasks such as loosening screws or nuts are still necessary because the parts of assembly products are connected by nuts or screws, but it is costly. Therefore, disassembly motion should be improved and analyzed for higher productivity. In assembly processes, analysis of manual work for efficiency improvement and human resource development had been conducted using motion capture and machine learning technology due to dealing with enormous motion data. This study proposes an analysis of disassembly tasks using machine learning with motion data. First, motion data for a disassembly task is obtained from optical motion capture which acquires time-series 3D positional coordinate data. Secondly, the classification before and after proficiency is identified by analysis of the motion data through machine learning. Finally, the results are discussed in terms of learning effect on proficiency classification and body part in the disassembly task.