Data Clustering and Prediction for Fault Detection and Diagnostics

The characterization of a data cluster representing a certain process behavior is achieved by developing steady-state nonlinear modeling of one or more critical signals as a function of other process variables in the system. This prediction model is used to detect either sensor maloperation or process anomaly by comparing the prediction and measurement of the same variable. A large database from a process control system can be grouped using clustering algorithms. Automated generation of prediction models are applied to an industrial process to study the performance of this database management approach.