Process Design for Robotic Unscrewing: Sensor-Based Anomaly Handling in Circular Production

Shorter product lifecycles increase raw-material consumption and intensify the need for effective end-of-life strategies to support a circular economy. While robotic assembly is well-established, automated end-of-life disassembly remains unresolved. Used products arrive in uncertain condition, introducing unpredictable anomalies, that current systems are unable to reliably handle. Unscrewing is a fundamental disassembly operation, yet highly susceptible to process anomalies. Conventional torque-based detection methods fail to reliably handle cases such as detached screw heads or occluded fasteners. This work contributes towards autonomous disassembly through a systematic redesign of anomaly detection for the robotic unscrewing process. Eight typical wear scenarios were identified from the fleet data of a power-tool manufacturer. Based on these scenarios, the unscrewing process was redesigned and a multimodal sensing system with calibrated accuracy requirements was developed. Anomaly detection is implemented using parameterised decision-tree models operating on torque, axial-force, and TCP-position streams. The system was validated on 3D-printed test specimens and on screws of a Fein CG15 angle grinder, achieving 100 \% sensitivity and accuracy on 410 trials across eight failure-mode classes. ... mehrDetected parameters can be persisted in an ontology-driven knowledge graph to enable cross-station parameter reuse.