A safety management approach for Industry 5.0′s human-centered manufacturing based on digital twin

Safety management is fundamental for ensuring human-centered manufacturing as defined by Industry 5.0, which requires the integration of knowledge-driven, human-machine-environmental safety. However, three challenges need to be addressed to fill the gap between contemporary workshop safety management and the expected requirement: insights into the complex interactions of human-machine-environmental activities, understanding the causality of unsafe states, and the adaptability of safety management methods. A reasoning approach towards factory unsafe states based on Digital Twin is proposed to address these challenges. First, a machine-readable semantic reasoning framework is introduced. Second, the ontology of unsafe states during production is modeled. Then, a high-fidelity virtual Digital Twin Workshop is constructed, which can simulate various workshop unsafe states and generate a virtual dataset. The virtual dataset is then mixed with the real dataset to train and test the target detection network, which is used to detect unsafe instances mapped to the ontology for reasoning. Finally, an experiment demonstrates that the proposed approach can address the three challenges.