On quick-change detection based on process adaptive modelling and identification

This work introduces a simple method and implementation algorithm for quick-change detection based on process measurements, adaptive modelling and identification, which will be called model-based quick-change detection (M-QCD). The idea of the method is to select a proper model for process identification purposes, and to continuously estimate the parameters of this model. A change in process implies a change in the set of parameters, which is detected using a criterion based on gradient and energy of the previous values of the criterion. The results are compared to those given by an off-line classical algorithm used in change detection area, based on variance. Computer simulations show better results for the proposed method and confirm that - some times - simple methods and algorithms, free of constraints, can work at the same performance level with change detection complex algorithms, which use advanced processing and optimization.