Abstract
Fault detection and diagnosis are extremely necessary in complex industrial systems. Data-driven monitoring technologies have been widely used to extract useful information from a large number of highly correlated process variables and historical data. Currently, the data-driven method is focus on Multivariate Statistical Process Monitoring (MSPM). The main characteristic of MSPM is to extract feature information from process data through dimensionality reduction algorithm, and to establish statistics for process monitoring.Thefaultdetectionbasedondataregressionorclassificationandthe fault diagnosis based on Bayesianinferencearediscussed here. These methods can exploit the underlying geometrical structure that contains both global and local information between the process variables and quality variables. So they can make the fault detection and diagnosis more accurate, more consistent, and faster.
