Statistical monitoring of control loops performance: an improved historical‐data benchmark index

Abstract

Control systems are key elements of virtually all industrial processes, whose performance directly impacts aspects as important as: product quality and variability, operations safety, process efficiency/costs and environmental impact. In this paper we address the problem of monitoring the performance of such control systems, and in particular a new historical-data benchmark index is proposed (I M ), which is able to discern between perturbations in the system's core modules, which are under the supervision of process owners, from those originated at the level of disturbances, usually involving other stakeholders. It is a generalization of the current index (I v ) as it can be shown that it reduces to this index for the particular case where the variability of the disturbances is the same as in the reference or benchmark period and in the monitoring period. The results obtained demonstrate that the proposed historicaldata benchmark index is able to maintain the target false alarm rate under situations where the variability of the disturbances increases, a situation where the current index, I v , fails. When the disturbances variability is maintained, both indices present similar detection capability, as expected. The subsequent identification of the modules in fault was also analysed, and the results show that the proposed methodology is able to identify the general source of the degradation in the controller performance, namely, if it is due to a perturbation within the system's core (and which loop is affected) or at the level of the disturbances (increasing variability of the loads or change in their dynamical behaviour).

Data-driven performance assessment of multivariable control loops using a modified Hurst exponent–based index

“…The advantage of the proposed approach over model-based indices, such as the MV index and the linear quadratic Gaussian (LQG), 3,21,26 is that it does not require any prior information on system parameters. Also, its advantage over historical data-based indices 1,22,27,28 (in which one period of previous operation data when the performance of the system is optimal is selected as benchmark data) is that it does not require optimal operation data of the system. Besides, the method presented by Das et al 25 also requires the optimal operation data of the system.…”