Non-linear principal components analysis with application to process faultdetection

Principal component analysis has been used for the development of process p erformance monitoring schemes for both continuous and batch industrial proc esses. However, it is a linear technique and in this respect it is not nece ssarily the most appropriate methodology for handling industrial problems w hich exhibit nonlinear behaviour. A nonlinear principal component analysis methodology based upon the input-training neural network is proposed for th e development of nonlinear process performance monitoring schemes. Kernel d ensity estimation is then used to der ne the action and warning limits, and a differential contribution plot is derived which is capable of identifyin g the potential source of process faults in nonlinear situations. Finally, the methodology is evaluated through the development of a process performan ce monitoring scheme for an industrial fluidized bed reactor.