Iterative Tuning of Linear Quadratic Controller for AMBs in a High Speed Compressor

This paper presents the design of an Active Magnetic Bearing (AMB) rotor radial levitation controller using an iterative tuning method developed for the Linear Quadratic Gaussian (LQG) design problem. An iterative method is used to find the ideal pair of state and input weights in the LQG objective function that result in the best performance and stability robustness characteristics of the closed-loop AMB system. A search method is used to identify a set of input weight and state weight pairs that result in AMB levitation controllers that satisfy the necessary requirements for digital implementation. Furthermore, the stability and performance requirements from the API 617 and ISO 14839 standards for new compressors with AMB support are included in the controller design process. Finally, the derived LQG rotor levitation controller is implemented on a high speed centrifugal compressor test rig with a rigid rotor. Experimental results are presented and compared to the measurements obtained with a preliminary PID controller. The measured unbalance response and rotor orbit plots demonstrate the benefits of the proposed LQG method.