KNOWLEDGE REPRESENTATION IN A GRAIN DRIER FUZZY-LOGIC CONTROLLER

This paper describes a method of using governing rules associated with fuzzy membership matrices to represent drier control knowledge in a f uzzy logic controller. The objectives of the controller were to obtain (1) outlet maize moisture content between 15.0 and 16.0% and (2) outl et maize breakage susceptibility as low as possible with a certain req uired drying rate. The governing rules contained information for contr ol decision-making including predicted moisture and breakage levels of dried maize which were derived from some measurable drying process va riables, current dryer operating conditions, and process disturbances. Fuzzy membership matrices consisted of state matrices and action matr ices. State matrices contained likelihoods of the process achieving co ntrol objectives at the current process state. Action matrices contain ed degrees of confidence of control actions in achieving control objec tives. All matrices were kept in five knowledge sub-bases along with t heir governing rules to represent drier control knowledge for differen t process states. A computer simulation showed that of 918 test cases, 793 control actions designated by the fuzzy logic controller matched preferred ones, that is, actions which matched human operators' prefer red actions, indicating that this method of knowledge representation i n a fuzzy logic controller was effective.