Stiction Diagnosis And Minimum Variance Performance Assessment Of Control System With Stiction

As the most commonly used actuators in the process industry,the control valve generally contain stiction nonlinearities,and the stiction often leads to the stable oscillation of the system,which reduces the production benefit.In this paper,the system with valve stiction is studied from two aspects:On the one hand,when the system oscillates,how to determine whether the source of the oscillation is stiction or something else,such as over tuning and external disturbances,is the key problem to be solved before making the system back to normal,however,most methods for diagnosing oscillation at present are aimed at the oscillation caused by a single cause,and there are few relevant studies for the situation of multiple causes existing at the same time in practice,and the criteria are not rigorous enough,especially for the determination of whether the stiction is one of the causes of oscillation.To solve these problems,this paper proposes a set of comprehensive oscillation diagnosis methods,including linear system oscillation diagnosis method based on autocorrelation function(ACF),Hammerstein model identification method based on grid search algorithm and nonlinear oscillation diagnosis method based on model.The proposed method can not only diagnose the oscillations caused by a variety of single causes in a single input single output(SISO)loop,but also diagnose the oscillations caused by multiple causes,and make a more accurate judgment on whether the stiction is one of the root causes.On the other hand,when the system oscillates due to stiction,it imposes additional constraints on the performance of the system in addition to the time delay.It is critical for the engineer who maintains the performance of the system to accurately determine the performance benchmark for the system with valve stiction nonlinearity.For the research of performance accessment of onlinear systems,most existing methods are need to meet the condition which the nonlinearity in the system is reversible or differentiable,however,the stiction is not reversible or differentiable;even for the methods which are not need to meet the above conditions and suitable for stiction nonlinearities,either large randomized trials are required,or additional performance limitations due to stiction are not taken into account.In order to overcome the above problems,this paper presents a performance accessment method for nonlinear systems with valve stiction.An approximate linear model of the valve is fitted by spectrum estimation,and then an autoregressive moving average model(ARMAX)with external input is used to identify the process and the model of disturbance.Then,an open-loop simulation is built using the estimated process model and the linear approximation model of the valve,Then,using the estimated object model and valve linear approximate model to establish open-loop simulation,combined with empirical mode decomposition(EMD),the minimum variance benchmark correction can be obtained;Finally,by using the minimum variance benchmark of the corresponding linear system calculated on the identified colored noise model and time delay,the minimum variance reference of the system with valve viscous nonlinearity can be obtained.With this method,a more accurate minimum variance benchmark including the performance limitation caused by stiction can be obtained without a lot of experiments,so as to make a more reasonable performance accessment of the system.