An empirical data driven based control loop performance assessment of multi-variate systems

In this paper, an alternative method for the assessment of multi-variate control loop performance without relying on any a priori knowledge of the interactor matrices is proposed. The performance of the control loop is calculated from data driven autoregressive moving average and prediction error model. It is observed that the limited data in scalar measure of covariance of predicted errors used for performance assessment results in incremental in initial part and tends to steady-state as time tends to infinity, but large number of samples gives risen in scalar measures and tends to infinity as time samples tends to infinity and therefore it becomes difficult to calculate the performance index. In this paper, the later problem is solved by considering initial part of scalar measures with steady value for next-to-next time samples to calculate the control-loop performance index which would be utilized to decide healthy working of the control loop. Simulation example is included to show the performance index of multi-variate control loop. The proposed method is compared with method available in the literature.