A digital twin-based method for production system performance prediction

Digital twin technology is increasingly applied to improve production systems performance due to the fast-growing artificial intelligent technology. By leveraging the real-time data-driven model, some intelligent services are developed for more effective management in manufacturing, such as real-time monitoring, productivity prediction, and production schedule. However, the reliability of data-driven models always relies on enough high-quality samples, which carries out challenges for modeling the whole production system digital twin, especially labor digital twins who are essential in the human-centric system. Therefore, to rebuild the whole lifecycle of the digital twin worker and then build a comprehensive production system digital twin, we propose a hybrid method to build the labor digital twin, which fuses a theory model and a data-driven model. When lacking valid samples to build data-driven models, theory models are used to replace algorithm models for predicting operational time and production system performance. Finally, we use an assembly production line as an example to evaluate the method.