Non-linear performance monitoring

Successful process performance monitoring depends upon the efficient and effective handling of plant data. Classical univariate statistical techniques are theoretically not capable of analysing process data that has been corrupted by measurement error, noise and where the variables exhibit collinear behaviour. Traditionally, univariate Statistical Process Control (SPC) systems only detect disturbances related to individual quality measurement sources, and as a result interactions between variables which are so important in complex processes are ignored. These limitations can be addressed through Multivariate Statistical Process Control (MSPC). Applications to two industrial processes are considered to demonstrate the power of multivariate non-linear performance monitoring.