Condition Monitoring of Control Loops

Abstract: The main concern of this work is the development of methodsfor automatic condition monitoring of control loops withapplication to the process industry. By condition monitoringboth detection and diagnosis of malfunctioning control loops isunderstood, using normal operating data and a minimum amount ofprocess knowledge.The use of indices for quantifying loop performance is dealtwith in the first part of the thesis. The starting point is anindex proposed by Harris (1989). This index has been modifiedin order to cover a larger range of processes. The same conceptis then used to assess the sampling rate in control loops.Other index-based monitoring methods where some amount ofprocess knowledge is available are discussed.The evaluation of the performance indices discussed requiresknowledge of the process dead-time. Therefore a concept calledevent-triggered estimation is introduced in the second part ofthe thesis. Both automatic data selection and dead-timeestimation methods are proposed and conditions for successfulestimation are discussed.The last part of the thesis deals with the diagnosis ofoscillations. A method to automatically diagnose staticfriction (stiction) in the actuator is presented. Furthermore,two methods are proposed which allow automatic distinction ofexternally and internally generated oscillations. All describedmethods have been implemented in a MATLAB-based graphical userinterface which is briefly described.