Abstract
This paper presents a learning method based on linguistic instructions using a qualitative linguistic model of the controlled process. The operator's linguistic instructions are interpreted as an adjustment variable, a direction of the adjustment and an information of linguistic hedge.And then the fuzzy control rules are properly adjusted. At the same time, another learning of surpressing the influence of the learning to the other manipulated variables is carried out in addition to the learning about the instructed manipulated variables. The qualitative linguistic model is introduced for the knowledge of this learning. This learning method is applied to the fuzzy control of a chemical reaction process. As a result of simulation, it is shown that the adjustment suited to the operator's control intention is accomplished. Next, a method of using control informations for learning is described, where the weight of the linguistic hedge included in the operator's instruction is changed according to the amount of the control difference. It is also shown by simulation that the 40% of the instruction frequency can be reduced by this method.
