PLS-based FDI of a Three-Tank laboratory system

The problems of fault detection and isolation of dynamic systems has been studied intensively in the recent years and many successful industrial applications have been reported. In the main these studies have been restricted to model based techniques, with few reports of successful implementation of data driven approaches. These data driven approaches have been range from the application of linear regression techniques, to neuro-fuzzy systems. This paper reports on application of, Multivariate Statistical Process Control (MSPC) methodologies, which can provide a diagnostic tool for the on-line or real time monitoring and detection of the process malfunction is proposed. Finally the effectiveness of Partial Least Squares (PLS) in FDI of the Three-Tank system are represented and discussed through simulation results.