Statistical Performance Monitoring Using State Space Modelling and Wavelet Analysis

This paper describes the application of multiresolution analysis to the states and the model residuals calculated through the application of canonical variate analysis (CVA) for process performance monitoring. Applying Hotelling T to the states and residuals and calculating the statistical limits will materialise in an excess of false alarms since the CVA model states and model residuals are observed to exhibit serial correlation. By applying wavelets to the states, the auto-correlation is removed and the standard monitoring metrics can then be applied. The performance and sensitivity of the proposed methodology was assessed for different monitoring indices using average run length (ARL) and false alarm rate as performance indicators. The basis of the study was a benchmark simulation of a continuous stirred tank reactor.