Detection and quantification of valve stiction by the method of unknown input estimation Saneej B. Chitralekha ∗ Sirish L. Shah ∗,⋆ J Prakash ∗∗ ∗ Department of Chemical and Materials Engineering,University of Alberta, Edmonton, T6G 2G6 Canada ∗∗ Madras Institute of Technology, Anna University, Chennai, 600044, India Abstract: Valve stiction is one of the most common causes of oscillations in industrial process control loops. Such oscillations can degrade the overall performance of the loop and eventually the final product quality. The detection and quantification of valve stiction in industrial process control loops is thus important. From previous studies in the literature, a sticky valve has been shown to have a distinct signature of the stiction phenomena in its valve positioner data. However, the position of the modulating control valves is seldom available. We consider the problem of estimating the valve position as an unknown input estimation problem. In this work, we propose a novel application of the unknown input estimator in order to estimate the valve position given a linear model of the process and closed loop input-output data. Using the estimated valve position data, we can detect and also quantify the amount of stiction. We demonstrate the efficacy of the method through simulation examples where a sticky valve is deliberately introduced in the closed loop using a two-parameter stiction model available in the literature. An industrial case study is also presented in which the algorithm accurately detects and quantifies stiction in a level control loop of a power plant. Keywords: Stiction;Valves;Actuator;Closed loop;Kalman filter. 1. INTRODUCTION The problem of stiction in valves is well known to be one of the primary causes of oscillations in industrial process control loops. Such oscillations can easily propagate to other control loops and degrade the overall closed loop performance of the process. The higher variability in the process variables due to such oscillatory loops will be re- flected in the final product in the form of larger variations in the product quality. The detection and quantification of valve stiction in such loops is the first step towards a diagnosis scheme in order to alleviate the effect of a sticky valve on the closed loop performance. A data based, non-invasive method which can automatically perform this task has been the main theme of stiction detection and quantification studies in the past literature. Horch [1999] presents a data based method which is based on the cross- correlation between the controller output and process out- put. This method, which can only detect stiction but not quantify it, cannot be applied to processes which have an integral action. In another recent study, Choudhury et al. [2006] presents a method for detecting and quantifying stiction for linear processes using the process output (pv ) and controller output (op ) data. This method is based on the fact that for a linear process under closed loop control, a sticky valve would induce nonlinearity in the pv and op signals and hence stiction can be detected based on the ⋆ Corresponding author. Email: sirish.shah@ualberta.ca (Sirish L. Shah) nonlinearity in the control error signal. The quantification part is based on the pv vs op plot, which is found to be elliptical in shape and the width of the ellipse is used to quantify stiction. This type of quantification is very useful since it expresses stiction as percentage of the valve travel which is the practice in the process industry for quantify- ing stiction. The disadvantage of this quantification is that the width of the ellipse will be dependent on the effect of loop dynamics (mainly controller tuning) on the pv. Therefore the estimated width of the ellipse is termed as ‘apparent stiction’. On the other hand if the valve position (mv ) data is available, then a simple plot of mv and op can be used to obtain the actual stiction. For a sticky valve the mv vs. op plot will follow well defined patterns depending on the type of stiction as shown in an earlier work by Choudhury et al. [2005]. The true stiction can be quantified using an appropriate width measure of this pattern. In Choudhury et al. [2006], the mv vs. op plot was not used for the quantification part since for most practical cases the valve positioner data (available only for ‘smart valves’) is not available. In the current work, we propose a novel method in which the valve position can be estimated by the method of unknown input estimation using a linear model of the process and the pv,op data available from the process. Since most of the control loops with a valve have linear controllers, usually PI(D)s, which are capable of giving satisfactory regulation, the assumption of linear process is reasonable. We apply the unknown input Proceedings of the 7th IFAC Symposium on Fault Detection, Supervision and Safety of Technical Processes Barcelona, Spain, June 30 - July 3, 2009 978-3-902661-46-3/09/$20.00 © 2009 IFAC 1432 10.3182/20090630-4-ES-2003.0103