PID based control performance assessment for rolling mills: A multiscale PCA approach

Abstract

This work is focused on exploring the use of Multi-scale Principal Component Analysis (MSPCA) for performance assessment and diagnosis of PID controllers in steel rolling processes. An optimal PID controller is used for performance benchmark. Diagnosis is implemented using an angle-classifier PCA-based approach. Theoretical framework of MSPCA, the performance index and the diagnosis procedure are developed. A procedure for implementing the proposed algorithm is presented. Simulations on the rolling mill roll force demonstrate reliable assessment and diagnosis results and practical use of the proposed method.