Preformance Assessment Of Control System: Methods And Application

Since Harries defined the variance of closed-loop outputs under a minimum variance controller (MVC) as controller performance criterion, lots of researches have been concerned with it and ignore if it is suitable for all kinds of controllers. Actually, for PID controllers, it's found that the MVC variance is unrealistic due to the structural limitation of PID controllers. In this context, an assessment index based on achievable minimum variance of PID controller is introduced, which could lead to a more accurate assessment of PID controller performance. To get access to the achievable minimum variance, ARMAX model of process is identified, along with the employment of SOSP (Squares of Sums Programming), which effectively solves the non-convex optimization problem related to minimization of output variance.Because there are kinds of disturbance in actual industry, it's not enough to indicate the performance of control systems just by only one index. Therefore, the influence of disturbance to control system performance is studied. The situation under abrupt change of random disturbance is especially analyzed and a two-model cumulative sum test is applied to diagnose the abrupt change. Besides, the cross correlation function of controller error and its increment was used to differentiate between the random disturbance and deterministic disturbance, and it's suggested different indexes for different kinds of disturbance.Control loop oscillating shows that the performance of control system decayed. This paper introduces how to monitor the oscillation quickly and proposes a comprehensive method with Oscillation Decay Ratio,Oscillation Period and Power Spectrum Distribution.What's more, diagnosing the reason is tried by experience.In the end, software for performance monitoring of controller system was developed with VB, OPC Specification and Access techniques. It's applied in Tennessee-Eastman process, monitoring lots of control loops online and showing comprehensively results, which is helpful for the engineer to diagnosis the control problem in time. Additionally, the development of the software and its experimental approaches might become an effective and promising way for practical design and implementation of control system performance assessment software.