(57b) Error Band Identification and Characterization for Advanced Process Controller (APC) Performance Assessment

This work addresses the characterization and analysis of closed-loop process data used to calculate controller performance metrics in a statistical process control (SPC) framework. SPC-based controller performance metrics address management's view of control systems as assets to be managed. The SPC approach utilizes a long-term (weeks or months) view of controller performance with a business emphasis on continuous quality improvement, identification of best practices, and the allocation of limited resources for control system maintenance. The SPC-based performance metrics in this paper are not used to identify controller tuning problems or valve stiction, but to provide opportunities for continuous process improvement of advanced process control systems by identifying ?assignable causes? for changes in controller performance. Controller actuating error (process variable value minus the set-point) is the key closed-loop variable for controller performance assessment when applied to feedback control. For a PID loop in an advanced process control (APC) system, the characteristics of the controller error are affected by factors other than those associated with the single-input-single-output (SISO) PID loop. The quality of the APC models, constraint specifications, and controller weight or priority settings will also impact the controller performance. The work reported in this paper is intended to facilitate methods to improve APC model maintenance and APC controller performance using historical process data available in the plant historian. Closed-loop archived data from a U.S. refinery are used in this paper to reveal the actuating error characteristics of multivariable controllers. Plots of the closed-loop data sets for the advanced process control manipulated variables (APC-MVs) exhibit different levels of variability when considered over a long period of time (one year). These periods of variability are termed as ?error variability bands.? Changes in the error variability bands are attributable to assignable causes responsible for changes in controller performance. Automatic identification of the error variability bands provides the starting point for further diagnosis and elimination of assignable causes that can lead to real business improvement. This paper presents four error variability band identification techniques using general purpose statistical tools including histograms, normal-probability plots, quantile-quantile plots and the sample autocorrelation function. The performance of these methods is presented using archived refinery data reconstructed on a one-minute sample period for flow, pressure, and temperature loops. The impact of set-point variability on APC manipulated variables is also illustrated. The variability in the actuating error is predicted from the process variable variability, set-point variability and the correlation between the process variable and the set-point and compared with the error variability estimated directly from the data. Although this paper is focused on APC manipulated variables, the results of this work are also applicable to the analysis of traditional, non-APC control systems.