| Research on developing automated controller performance monitoring systems has been increasing in the past decade. Control strategies such as Model Predictive Control (MPC) or other supervisory control are crucial for optimization of process operations. Performance gains from such advanced control techniques diminish if the lowest element in the control hierarchy is ineffective in tracking (or rejecting disturbances) as dictated by higher-level controllers. Deterioration of control performance may have several reasons such as: badly tuned controllers, oscillating load disturbances or nonlinearity in actuators. The pneumatic control valves that are the lowest elements in the control hierarchy are often implicated in control performance degradation; accounting for about 32% of control loops.;Most of the solutions that have been proposed for this problem are based on multivariate analysis. In multivariate analysis, in many cases, arbitrary assumptions need to be made to identify the root cause for oscillation. In this work, it is proposed to move away from multivariate analysis and categorize the root cause for oscillation in each individual loop as being internal or external to the loop. Such an analysis on all the loops will lead to the identification of the root cause for oscillation. Also, once the problematic control loop is identified, the proposed approach should point out the ultimate cause. The ultimate cause might be bad tuning, or stiction or combination of these along with or without external disturbance. The proposed approach provides a framework to comprehensively identify all these failures that are possible.;Keywords. root cause analysis; valve stiction; stability; Hilbert-Huang Transform; Non linear process; model identification and analysis.;Most of the automated controller performances monitoring approaches identify all the control loops that are poorly performing. This identification is done based on whether the observed oscillations are within acceptable limits or not. From this group, a further set of controllers with unacceptable oscillations are identified. However, identification of the control loops that are a root cause for the observed oscillations is a non-trivial task. This is mainly because of the interactive nature of multivariable systems. Current controller performance monitoring approaches offer semi-manual procedures for identifying the loops that are the root cause for oscillations. Manual procedures are expensive, time-consuming and in many cases impractical. Hence a fully automated approach for root cause analysis of oscillating control loops is of immense practical interest. |
