Neurofuzzy model based predictive control for thermal batch processes

In many cases, it is difficult to derive a precise mathematical model, based on first principles, for a given process. Besides, the computation of the solution of models obtained through this methodology may require a large computational effort making them useless for real time tasks like control or optimization. Neurofuzzy modelling, which permits an easy way to derive successful models, is a good alternative which can be employed to overcome such limitations.In this paper, together with the neurofuzzy modelling, several strategies based on non-linear predictive control are presented. The low computational cost associated with neurofuzzy models and controllers makes them suitable candidates to be implemented into industrial Programmable Logic Controllers (PLC). Both the model and controllers are validated and implemented in a pilot plant for the thermal sterilization of solid canned food in steam retorts and based on the results, a comparison between the different predictive control strategies is presented.