Dynamic human reliability and dual-type quality prediction in human-machine collaborative tuning manufacturing systems

In human-machine collaborative manufacturing, dynamic human reliability (learning, forgetting, fatigue, and recovery) critically affects product quality, especially in complex debugging processes. This paper proposes a comprehensive analytical framework for human reliability and quality defect prediction in human-machine collaborative tuning. First, a dynamic human reliability model integrating the dual mechanisms of ”learning-forgetting” and ”fatigue-recovery” is constructed to quantify the evolution of the Human Error Probability (HEP) over time. Second, based on the time-varying HEP, two types of quality defect prediction models are established: a Non-Homogeneous Poisson Process (NHPP) for discrete hard failures, and a quality characteristic distribution drift model for continuous soft failures. Furthermore, a Decision Regression Tree (DRT) algorithm is introduced to predict critical human reliability parameters from multi-dimensional manufacturing influencing factors. Finally, the methodology is validated through an application to the tuning process of electromechanical relays. The results demonstrate that the framework effectively captures dynamic operator reliability changes and accurately predicts both discrete and continuous quality defects, offering a new theoretical basis and analytical tool for active quality control and human factor risk management.