Digital Twins (DTs) developed through simulation models have become essential tools for monitoring, optimisation and decision-making in industrial operations. However, as these models remain active throughout the production lifecycle, their validity may deteriorate due to changes in the physical environment, requiring robust, interpretable and ongoing assessment. This study proposes a statistically grounded and diagnostically oriented approach based on multivariate control charts, specifically Hotelling’s T² chart, to evaluate the lifecycle validity of simulation-based DTs in automated production settings. The method not only detects statistical deviations but also enables root cause diagnosis and guides decisions on model intervention or continued use. A real-world application is presented involving an automated manufacturing cell, where data were collected via industrial sensors and analysed in near real time, in this study, updated daily using data from the previous day, although the method supports more frequent or continuous execution. The results demonstrate the method’s ability to identify deviations, isolate contributing variables through variable-wise decomposition, and reinforce trust in DT models during operation. By addressing the validation of DTs as a continuous and multivariate challenge, the proposed approach enhances existing practices and offers a structured, practical solution for real industrial systems.
Lúcio et al. (Tue,) studied this question.