Online validation of digital twins for manufacturing systems

Abstract

Digital twins are considered among the most important technologies to optimize production systems and support decision making. To benefit from their functionalities, it is essential to guarantee a correct alignment between the physical system and the associated digital model, as well as to assess the validity of the digital model online. This operation should be conducted rapidly and with a small data set, especially in highly dynamic contexts. Further, the whole behaviour of a system may be of interest rather than the sole average performance. Traditional validation techniques are limited because of the restrictive assumptions and the need for large amounts of data. This work defines the problem of checking the validity of digital twins for production planning and control while the physical system is operating. A methodology describing the data and the types of validation is proposed including a set of techniques to be used at different levels of detail. The congruence between the physical system and the corresponding digital model is measured by treating data as sequences and measuring their similarity level with digitally-produced data by exploiting a proper comparison technique. The numerical experiments show the potential of the proposed approach and its applicability in realistic settings.