On-Line Noise Estimation for Adaptive SPC Loop Control

Abstract

This paper describes a method of measuring the variance of a noisy process on-line, enabling the application of adaptive Statistical Process Control (SPC) to supervise a continuous process loop. A fundamental requirement of SPC techniques is that the data are independent. Process control loops typically generate autocorrelated data, and this tends to create an increase in the false alarm rate when attempting to apply SPC techniques. The method presented here produces an uncorrected sequence from potentially correlated data, overcoming this problem. The SPC supervisory unit acts externally to the control loop preventing any lags in feedback. When these SPC charts are used in a supervisory role with a conventional PID controller, application of individual PID terms achieves a much quieter control with little reduction in performance and without the need for prior off-line noise tests. The results of tests in simulation and on a steam-water heat exchanger are presented.