On-line adaptive fuzzy control of a cryostat

One of the reasons for the current interest in fuzzy control is the successful application of simple rule-based controllers in many diverse fields. However, relatively few adaptive neuro-fuzzy schemes have been tested on practical problems. Moreover, many strategies based on neuro-fuzzy techniques are not suitable for dedicated implementation, since they use some form of optimisation (e.g. recursive least squares or direct search techniques) to select the required control action or for training. To fill this void, the performance of an adaptive fuzzy control scheme, which combines a simple fuzzy identification algorithm with the feedback error learning method, is assessed by using it to control the temperature in a liquid helium cryostat. Results from the practical experiments are presented in this paper.