Control Loop Performance Monitoring via Model Validation
Over the last twenty years, research on the detection of abrupt changes has emerged as an important area in the control community owing to the pioneering work of Basseville and Nikiforov (1993) and references therein. The problems concerned with this subject have focused, in the main, on fault detection and diagnosis, data segmentation, gain updating for adaptive algorithms, model validation and process quality control. Among various statistical-based detection algorithms, the local approach has recently regained significant interest owing to the notable work of Basseville (1998) and Zhang et al. (1998). The effectiveness and reliability of this approach has been demonstrated by its application in the monitoring of critical processes such as nuclear power plants, gas turbines, catalytic converter etc.(Basseville 1998). The local approach has a number of distinct features, among which are sensitivity to small changes and simplicity with the asymptotically uniformly most power-fulness. In addition, local approach algorithms can often be developed along the same lines as model identification. Thus, many established methods to enhance identification algorithms may be transfered readily to local detection algorithms.
Unable to display preview. Download preview PDF.