Automated Detection of Disassembly States in Electric Motors Using Point Cloud Comparison

The depletion of natural resources and mounting environmental degradation necessitate a fundamental shift towards a long term sustainable economic system. Automated disassembly is an important part of innovative end-of-life solutions on the way to a sustainable circular economy. Disassembly sequences and associated processes can be planned on the basis of suitable digital representations of the object to be disassembled. However, due to the inherent uncertainty of used products, deviations between the digital representation and reality have to be expected. This means, that an ideal execution of the planned disassembly sequences cannot be assumed and individual processes might fail. A truly automated disassembly system therefore needs to be able to detect deviations between digital representation and reality, both for the disassembly object in itself as well as the expected and actual process outcome. This paper presents a system designed to determine the true disassembly state of electric motors by matching a geometric motor model from a disassembly simulation with reality. This is achieved by comparing point clouds generated from the disassembly simulation to those captured by a 3D camera system. Additionally, it is explored how contextual information, such as the movability of individual parts, can be utilized in the matching process. The system's effectiveness was validated with a simplified dummy motor.