Diagnosis of Closed Control Loop Status Using Performance Analysis Based Approach

Abstract

A novel approach for systematically assessing the current feedback controller performance, diagnosing, and removing the root causes resulting in performance deterioration is proposed. This fault diagnosis methodology does not require the traditional complex physical model and/ or a prior external input to perturb the operating system in order to achieve accurate fault identification. This is achieved by a series of the statistical hypothesis procedure, testing an impulse response of the control output which includes the feedback invariant (FBI) and the feedback dependent (FBD) terms. Both term variances via the statistic hypothesis testing are performed by comparing the current control and benchmark operating conditions. If the current performance has significant discrepancy from the benchmarked one, then the just-identified fault element (the process dead time, the controller tuning, the process, or the unmeasured disturbance) is isolated first and then the same diagnostic procedure based on FBI and FBD for exploring the new possible fault is successively applied. According to the characteristics of the closed loop output response, a diagnosis tree structure that has the characters of hierarchy and integrated knowledge is established. It is organized based on the sequence from general to the special. The capability of the proposed method is illustrated through a simulation case and a level tank system, including the single and the multiple fault problems.

Diagnosis of Cascade Control Loop Status Using Performance Analysis Based Approach

“…18,19 Using the routine operating data to construct a closed-loop impulse response model, Yea and Chen developed a diagnostic reasoning tree that could locate and remove root causes of poor performance in a feedback control system. 20 This article is an extension of our recent work, 20 and it presents a diagnosis methodology for cascade control loops based on control performance. Like the single loop of feedback control, the primary and secondary output variances can be separated into cascade-invariant (CI) and cascade-dependent (CD) terms, respectively.…”

“…With prior process knowledge, the data-driven method was used to explore the root cause. , Choudhury et al generated graphical plots by higher-order statistical techniques to quantify non-Gaussanity and nonlinearity and diagnosed the cause of poor performance of the control loop . Under the IMC structure, Cinar et al and Stanfelj et al monitored the feedback and feedforward systems with a small perturbation at the setpoint via a decision tree. , Using the routine operating data to construct a closed-loop impulse response model, Yea and Chen developed a diagnostic reasoning tree that could locate and remove root causes of poor performance in a feedback control system …”

“…This article is an extension of our recent work, and it presents a diagnosis methodology for cascade control loops based on control performance. Like the single loop of feedback control, the primary and secondary output variances can be separated into cascade-invariant (CI) and cascade-dependent (CD) terms, respectively.…”

Contribution to the study and design of advanced controllers : application to smelting furnaces

“…The stochastic based index most commonly studied is the minimum variance controller calculation. This index is based on the variance of the process output, which is compared as a benchmark to the minimal theoretically achievable variance control (MVC) Harris [1989], Yea and Chen [2005]. In turn, deterministic methods include classic performance measures (i.e.…”