Neural-fuzzy Model versus Multi-model based Controllers Applied in a Water Gas Heater System

J. Vieira and A. Mota (Portugal)


Neuro-Fuzzy Control, Multi-Model Control, Switching Methods, Model Based Control, Prediction.


This paper presents the control of a water gas heater with two Smith predictor based model controllers. The modelling types used are neuro-fuzzy global modelling and multi-local-modelling. The system is a domestic water gas heater that presents a varying time-delay and measured perturbations like water flow, cold water temperature and wanted output water temperature variations. The main objective is to control the output water temperature comparing the characteristics of algorithms, execution times and real control results. It presents a detailed description of the water gas heater in terms of its physical characteristics and laws. Direct and inverse neuro-fuzzy global models are presented, as well as, the implementation description of the local linear sub-models and the respective new switching method used. The two control strategies, neuro fuzzy Smith predictive controller and the multi-model Smith predictive controller, are tested in the real system. The real time results, advantage and disadvantage of each type model/controller are presented.

Important Links:

Go Back