Numerous analytical methods have been proposed to monitor and evaluate control loops. Newer approaches compare controller performance against a combination of performance measures to produce an overall performance index. While these performance measures offer potential benefit to plant operation, they are not easily interpretable by industrial practitioners, and thus are of limited value. For an analysis to be useful, the communication must be clear and concise. A performance index must alert operations staff when performance is degrading, and must provide specific information about the cause of the changes. [1]

To provide a more industrial-friendly performance monitoring tool, this research considers traditional statistics in a new way. Moving average, moving variance/standard deviation, autocorrelation, cross-correlation, and power spectrum are the statistics focused on in this study. Results presented show that novel graphical representation of these statistics reveals process events, including declining performance, drift, periodic behavior hidden in process data, changes in the nature of disturbances, and interactions between specific process elements.

Computational and exploratory data analysis methods were applied to identify patterns in multivariate data sets. The statistical investigative methods include such techniques as examining distributions of variables, monitoring coefficients against certain thresholds, identifying and classifying peak locations, and differentiating curvature variations. [2-6]

Using data collected from an industrial material processing operation as well as simulations of various common chemical processes, this work employs these pattern recognition techniques to characterize process performance. A systematic strategy is developed to automate process analysis and facilitate problem resolution.

Rules for investigating process behavior are established, including guidelines for when and how to apply the statistics in addition to how to methodically interpret the results. Specific graph elements that assist in diagnosing process behavior will be presented. Example processes include detailed simulations and bench-scale experiments.

[1] “An Overview of Control Performance Assessment Technology and Industrial Applications”, Mohieddine Jelali, Control Eng. Practice, 14, 441-466, (2006) [2] "On-line Pattern-Based Part Quality Monitoring of the Injection Molding Process" Suzanne Woll, B. Souder and D. J. Cooper, Polymer Engineering and Science, 36, 1477 (1996) [3] "A Unified Excitation and Performance Diagnostic Adaptive Control Framework" Ralph F. Hinde, Jr. and D. J. Cooper, AIChE Journal, 41, 110 (1995) [4] "Using Pattern Recognition in Controller Adaptation and Performance Evaluation" Ralph F. Hinde, Jr. and D. J. Cooper, Proc. 1993 American Control Conf., IEEE Publications, NJ, 74 (1993) [5] "Neural Network Based Adaptive Control via Temporal Pattern Recognition" Lawrence Megan and D. J. Cooper Canadian Journal of Chemical Engineering, 70, 1208 (1992) [6] “Statistical Pattern Recognition: A Review” A. K. Jain, R. Duin, J. Mao, IEEE Transactions on Pattern Analysis and Machine Intelligence, 22, 1, 4-37 (2000)