Computationally efficient implementation of robust controllers in active magnetic bearing systems

Abstract

This paper proposes a novel method for implementing robust controllers for active magnetic bearing (AMB) systems in a computationally cost-effective manner without affecting robust performance. The method decomposes the single-rate LTI controller into two parts, one containing slow modes and the other fast modes, where the slow and fast modes refer to the low and high-frequency modes, respectively, relative to a threshold frequency. Slow modes of the controller do not need to be implemented at a fast rate to achieve an accurate response and, therefore, can be implemented at a slow rate. This transforms the single-rate LTI controller into a dual-rate periodically time-varying controller and can significantly reduce the computational cost of necessary algebra at each time step using the interlacing technique. The novel part of the method is the selection of the dual-rate configuration, i.e., selection of the threshold frequency for the decomposition and the implementation rate of the slow modes. The proposed method utilizes the largest singular values of the closed-loop system along with μ-analysis and lifting technique to analyze robust performance of different dual-rate implementations and determines the optimal dual-rate configuration that maintains robust performance while maximizing computational savings. The presented method is demonstrated experimentally on an AMB test stand. A dual-rate configuration is identified that reduces the computational cost of the necessary controller algebra by 35% while maintaining robust performance. Experimental comparison of the performance and computational cost of the single-rate and the dual-rate implementations of the controller are shown. The results demonstrate the feasibility and potential of the presented method.