Hybrid fuzzy modeling of chemical processes (Q1855494)

The expert knowledge, local linear models, and historical data are combined into a uniform ``grey-box'' fuzzy modeling framework. The expert fuzzy model (EFM), local linear models (LLM), and the compensation fuzzy model (CFM) constitute the basic layer. At the second layer the special module decides which model in the basic layer should be used to make the final prediction. Its fuzzy selection rules are compatible with the understanding of engineers. If there is enough training data and the CFM is reliable then use the CFM. If the CFM is not reliable but the EFM can provide some reliable compensation, use these two models to form the hybrid fuzzy model output. If neither the EFM nor the CFM give reliable predictions, an LLM should be used as a default model. The proposed hybrid fuzzy modeling framework is applied to a simulated pH neutralization process.