A predictive fuzzy relational controller

The fuzzy controller proposed in this paper has the same structure as a conventional model-based predictive controller but uses a relational-based model which is superior to rule-based models in that it: permits numerical and analytical analysis; employs the max-product compositional operator which gives better results, using a minimum distance criterion than the traditional max-min compositional operator; and includes self learning (fuzzy identification) capabilities. Simulation results show that this fuzzy controller gives better performance than a gain-scheduled PI (proportional plus integral) controller when applied over the full operating range of a continuous nonlinear process.