On the Implementation of Adaptive Inverse Control To Virtual Transfer Systems

The potential of real-time hybrid simulation (RTHS) has been increasingly explored over the last twenty years, bringing numerous methods into focus, both from a scientific and a technological perspective. In contrast to other forms of hybrid simulation, such as pseudodynamic substructuring, RTHS poses significant challenges in both its numerical and experimental counterparts, mainly due to the hard real-time constraints that must be met during each execution of the associated control loop. As far as the numerical substructure is concerned, the time step of the numerical integration scheme should conform to the step of the real time control loop. This is quite demanding for substructures of increasing complexity and usually necessitates the introduction of a reduced order model [1]. The numerical substructure is linked to the physical one through a transfer system, which brings additional complexity into the loop, as it is characterized by its own dynamics. Here, issues involving the accurate reproduction of the control signal, the elimination of the inherited time delay, the control-structure interaction and the rejection of all disturbance sources, are still subject to intensive research. Among other considerations, the application of adaptive inverse control (AIC) methods [2] has shown quite good performance in terms of accurate time-series reproduction. In this study, we explore further the effectiveness of AIC via its implementation to a recently published virtual RTHS benchmark problem [3]. In specific, we adopt a filtered-X AIC framework and we report on its effectiveness in modelling the control plant (e.g. transfer system and physical substructure) under representative input signals, partitioning schemes and parametric uncertainties. Our assessment shows a high degree of efficacy and robustness for the proposed AIC architecture and suggests further investigation, mainly in actual RTHS tests