| Scope and method of study. The main objectives of this study were to improve the process of selecting the valid rules and to accurately predict the value of the output. A hot and cold water mixer simulator was used to generate the data. The For the purpose of comparison, the truth calculations for the antecedent and consequent were performed using both the geometric operator and minimum operator. For each operator, different rule sets were obtained by using different threshold values for corroboration and merit. A statistical technique was developed to quantify the delay. In the prediction phase, Kumar's modified prediction technique was compared to the conventional prediction technique used in fuzzy logic. The RMS prediction error obtained from these techniques were calculated and compared.; Findings and conclusions. The minimum operator was found to be more suitable, for performing the Ta and Tc calculations, as compared to the geometric operator. Corroboration was sufficient to assess the validity of the rules and merit was not required. The threshold value of corroboration equal to unity has been proposed in this work. In addition, it has been proposed that only one out of all the possible interacting rules should be included in the final rule set. In general, conventional prediction technique was found to more accurate and computationally efficient then Kumar's modified prediction technique. In addition, minimum RMS errors were obtained when the value of the number of selected (valid) rules was close to the minimum number of rules required to describe the system completely. |
