Model-free parameter tuning with continuous genetic algorithm: Application to virtual controlled object-based active vibration controller

Abstract

Active vibration control is a powerful approach to provide higher damping effects over a broad range of frequencies. However, most of them generally involve a modeling process of a controlled structure, complicating the controller design. The main contribution of this study is to present a novel tuning scheme of a model-free active oscillation controller based on continuous genetic algorithm (CGA), which can simplify a controller design process by eliminating the need for plant modeling. The model-free controller is developed using a virtual controlled object (VCO) as its foundation. VCO makes it possible to construct an active damping controller without knowing any specifications of a real controlled structure. In the offline tuning of the design variables of the controller, we use a reference-controlled object (RCO) - a single-degree-of-freedom (SDOF) structure defined by the designer - to assess the damping level instead of the actual controlled structure model. This evaluation is feasible due to the robustness of the VCO-based design, which accommodates variations in the controlled object structures. The application of the CGA, which has the global search ability, enables automatically specifying the optimal parameter values without costly trial-and-error works. The effectiveness of the auto-optimized model-free damping controller is demonstrated through numerical examples across different mechanical systems, with comparisons made to a trial-and-error-tuned controller. The main finding is that the controller can reduce the vibration amplitudes even though plant modeling of the actual controlled objects is not accompanied at all. As the representative results, the resonance peaks of two different cantilever plates were suppressed by 14.38 and 27.13 dB in the low-frequency band. The detailed-specification-independent nature provided by the model-free design makes it possible to apply the resultant controller to various structures because the robustness is enforced.