An Intelligent Model Based on TS NARX for Process Prediction and Diagnosis Rule Extraction

In this paper, an intelligent model constructed with fuzzy TS dynamic nonlinear autoregressive with exogenous input (NARX) is introduced for process state identification and behavior prediction for complex processes. In the model, fuzzy neural networks (FNNs) are applied as process state classifiers for process state (fault) detection. An optimization schemes are also investigated for model adaptability to cover time depending process changes. After model optimization, the process difference process state and its input data state can be determined based on the classified process state and input variables. Data mining is employed to discover valuable knowledge and rules hided in process data. Finally, a real case is studied for products supply process diagnosis and forecasting with this model. It indicates that the model has good performance for process state classification, identification and process behaviors prediction, as well as business rules extraction for making decision.