(654f) Diagnosis of Source of Oscillations in Linear Closed-Loop Chemical Processes | AIChE

(654f) Diagnosis of Source of Oscillations in Linear Closed-Loop Chemical Processes


Srinivasan, B. - Presenter, Columbia University
Nallasivam, U. - Presenter, Clarkson University
Rengaswamy, R. - Presenter, Texas Tech University

In a single-input single-output (SISO) closed-loop system, under constant or non-oscillatory set-point, oscillations in the output can occur mainly due to one or many of the following reasons: (i) aggressive tuned controller, (ii) presence of stiction in control valve and (iii) disturbances external to the loop. Oscillations in control loops increase variability in product quality accelerate equipment wear and may cause other issues that could potentially disrupt the operation. Therefore, it is important to identify the source of oscillations in the control loops. Typically, in industries, the diagnosis algorithm for identifying the sources of oscillation in control loops should meet the following conditions: (i) the algorithm should perform well just with the available controller output (OP) and process output data (PV). No external excitation is allowed and (ii) the algorithm must work without any prior model information. With these requirements, it is a challenging task to detect the cause of oscillations in linear closed-loop systems.

The proposed algorithm for oscillation diagnosis in linear processes is designed to satisfy these realistic requirements with minimal assumptions (realistic) regarding the disturbance and noise corrupting the process. The key aspects of the proposed method are: (i) it combines both parametric Hammerstein algorithm (stiction identification algorithm) and a non-parametric (Hilbert-Huang transform HHT) technique (used to distinguish controller and external disturbance) for diagnosis of source of oscillations and (ii) identification of unique frequency domain signature for diagnosis of disturbance and controller related oscillations. Results obtained from simulation and industrial chemical process case studies demonstrate the utility of the proposed method for oscillation diagnosis in closed-loop systems.