Differential evolution approach for tuning an H∞controller in AMB systems

In this paper a controller for an uncertain active magnetic bearing system is synthesized. A loop-shaping H_∞ approach is applied and the performance weights are selected to fulfill the frequency domain specifications. Weight selection is usually done from the engineering perspective and needs some experience. In this work, the selection is automated with a differential evolution algorithm. Objective values and constraints based on singular values of the system are applied. Single-and multiobjective algorithms are evaluated from the viewpoints of robustness and performance. The results show that the differential algorithm can improve all the objectives compared with a system tuned by a human operator.