Fault Detection in a Continuous Pulp Digester

Diagnostic strategies for Fault Detection and Isolation in a continuous pulp digester are presented. Several methodologies for Fault Detection and Isolation are compared for accuracy and real-time implementation potential. A Gross Error Detection methodology for biased and noisy measurements is initially examined. Next, a Gaussian Radial Basis Function neural network approach for detection of product quality related changes is also used. Changes in feedstock composition using an additional neural network approach is also developed. The efficiency and limitations of all methods are demonstrated using a first principles model.