Valve Stiction Quantification: a Robust Methodology to Face Most Common Causes of Loop Perturbations

Abstract

Valve stiction is one of the most common causes of poor performance in control loops. Quantification of valve stiction is of crucial importance for maintenance scheduling, but can be heavily affected by the presence of unavoidable perturbations in loop variables. A procedure which allows stiction modelling and quantification is proposed. The technique requires only data normally registered in industrial plants - controller output and controller variable - and permits to reproduce the unknown real stem position. The method describes the system by means of a Hammerstein model: a non-linear block for valve stiction followed by a linear block for process dynamic. A two-parameters empirical model is used to reproduce accurately the valve behaviour and a linear least square identification is performed. A grid search method allows to estimate process and stiction parameters. The paper analyses how several loop perturbations as setpoint variations, controller tuning and external disturbances, could affect the stiction estimation obtained by this inherently robust technique. A general methodology is proposed to discard data for which quantification is very likely to give wrong indications and to restrict the application to appropriate cases. Simulations show that several sources of perturbations can be eliminated, thus improving the reliability of stiction evaluation. Results are confirmed by the application of the method to industrial data.