Real-time hybride-fysiek-virtueel testen in de industriële praktijk brengen

Real-Time Hybrid-Physical-Virtual Testing (HPVT) enables full-system testing by combining physical prototypes with simulation models in real-time, offering significant advantages for industrial product development: reduced testing costs, accelerated development cycles, safe testing of high-risk scenarios, and early-stage system validation. HPVT relies on recent technical advancements in model-based techniques such as hybrid simulation, x-in-the-loop testing, and co-simulation. However, widespread industrial adoption of these advancements is still at an early stage. This thesis aims at bringing real-time HPVT methods into industrial practice by tailoring and incorporating them into industrial real-time solutions. This should result in an efficient, performant and scalable way of applying real-time HPVT to industrial use-cases in sectors such as the automotive and aerospace ones. Two fundamental challenges are addressed: monitoring and improving the HPVT performance, to achieve robust stability while balancing accuracy and computational performance. For performance monitoring, energy-based coupling performance indicators (residual power and energy) are investigated for industrial HPVT applications using an originally-designed e-motor test bench. An extensive error analysis reveals how these indicators are influenced by subsystem dynamics, sensor characteristics, coupling parameters, and coupling faults introduced through systematic fault injection. Results demonstrate that residual power and energy serve as robust coupling performance indicators for HPVT systems. The investigated capabilities to detect both gradual parameter changes and sudden system faults opens new possibilities for real-time monitoring and adaptive control strategies in fail-safe HPVT systems. For performance improvement, Iterative Learning Control (ILC) methodologies are developed to compensate for errors at two levels: coupling delays between subsystems and actuator dynamics within subsystems. For coupling-level compensation, two similar ILC approaches are developed to compensate for coupling delays. They are validated on both fully simulated and HPVT setups, showing that the coupling error is reduced with several orders of magnitude. However, non-repeating noise present in the HPVT setup hinders ILC convergence, and filtering countermeasures have to be implemented. For subsystem-level compensation, again two ILC approaches are developed that compensate for HPVT actuator dynamics. They are applied to a representative RTHS benchmark. The first approach involves model-inversion ILC combined with feedback control, to cancel out the dynamics introduced by the actuator and system-under-test. The second approach revolves around a data-driven adaptive inverse control approach that achieves near-perfect tracking, except for an introduced time delay. Here, ILC is applied to compensate for this artificial time delay. A comparative analysis of all ILC-based methodologies establishes necessary conditions for successful ILC application in industrial HPVT: system stability or stabilisation via feedback control, and reasonable model assumptions, such that robust performance and fast convergence can be achieved in the presence of iteration-varying references. The developed methodologies are real-time compatible and require minimal subsystem information, making them suitable for industrial environments. To conclude, this research provides practical solutions for industrial HPVT deployment, demonstrated and validated for representative applications covering co-simulation, RTHS structural testing, and automotive electric powertrain testing. The findings pave the way for wider adoption of HPVT in industry, enabling more efficient and reliable product development processes.