Offline digital twin synchronization using measurement data and machine learning methods

Digital Twins play an important role in modeling production processes to adapt parameters according to predicted situations. Panel bending machines from Salvagnini use this technology to ensure safe operating conditions and to guarantee accurate results for different settings, even with highly variable material properties. Due to constantly increasing accuracy requirements, digital twins have to increase accuracy on the one hand and adapt to new machine generations on the other hand. This work shows how machine learning tools can be used to synchronize digital twins accurately and efficiently with real-world behavior by learning parameter values with measurement data while maintaining interpretable and robust analytical models.