MONITORING THE VALIDITY OF SIMULATION-BASED DIGITAL TWINS: A MULTIVARIATE APPROACH

Simulation-based Digital Twins (DGs) have played a central role in the digitalization of industrial processes, enabling predictive analytics and real-time decision support. However, the validity of these models tends to deteriorate over time, requiring continuous monitoring methods. This study proposes a multivariate approach to evaluate the adherence between the digital model and the physical system, using the Hotelling T² chart applied to the differences between real and simulated data. The proposal was tested in a real case with four monitored variables. The analysis allowed the identification of specific deviations that could go unnoticed in univariate approaches. The method proved to be effective in signaling inconsistencies and supporting decisions on model revalidation. By treating validity as a dynamic aspect, the approach contributes to the robustness of DGs in continuous operation in industrial scenarios.