Sigma-point Kalman Filtering in Physics-based Digital Twin Applications: Synchronization between Simulation and RealWorld

Abstract

In recent years, the integration of real-world systems and physics-based simulations has become increasingly necessary. Multibody-based simulations, in combination with information-fusing techniques such as nonlinear Kalman filters, have enabled the development of digital twins that can operate concurrently with physical machines. This parallel operation provides valuable insights into machine operation, supporting the transition to digital twins, digitalization, and green transformation.

This dissertation presents novel techniques for enhancing the performance of nonlinear Kalman filters, specifically in the context of hydraulically actuated machines. These techniques enable the estimation of unmeasured quantities, or virtual sensing, as well as the improvement of measured data quality for real-world systems. Additionally, adaptive techniques are introduced that provide estimations of noise statistics that affect the system and its sensors, in addition to estimating system states.

Considering the importance of computational complexity in achieving real-time simulation and minimizing computing resource costs, a new Kalman filter is proposed that requires significantly fewer computing resources than conventional Kalman filters. This dissertation explores various Kalman filtering techniques, providing analysis and comparison of academic and industrial examples. Overall, these contributions provide significant advancements in the development of techniques for enhancing the performance of nonlinear Kalman filters, supporting the use of digital twins in the advancement of real-world systems.