Joint online validation and parameter calibration within a digital twin framework: an approach based on bootstrap particle filtering

Manufacturing systems are increasingly adopting digital twins to enhance production performance and enable real-time decision-making. A critical challenge is maintaining the virtual model’s fidelity over time, as model drift can occur due to changing conditions. This paper proposes an approach for online validation and online calibration of manufacturing system digital twins in which the virtual part is a discrete event simulation model. The approach continuously compares event sequences and performance indicators between the real system and its virtual counterpart, detects discrepancies, diagnoses their root causes, and updates model parameters automatically to sustain alignment. A lab-scale production line is used for the implementation tests as a proof-of-concept. Experimental results demonstrate that the method identifies model divergence and adjusts the twin, maintaining accuracy in real time. This work contributes to building an integrated validation–calibration methodology that addresses parametric model drift, providing a pathway to more resilient and self-adaptive digital twins for manufacturing systems.