An Approximate Inference Scheme Based on Distance-Induced Relations: Weld Forming Prediction

Abstract

The construction of fuzzy conditional relations has been considered in Raha’s similarity-based inference schema, but inappropriate conclusions will be drawn in some cases due to the use of a single modification function. This paper presents an approximate reasoning method to generate induced conditional relations from distance measures. Two types of modification functions are given, and the type can be chosen according to the inclusion relation between a given fact and the antecedent of a rule. The fuzzy logic operators for constructing conditional relations are discussed, and the monotonicity and approximation properties of the reasoning scheme are inspected. By comparing and analyzing some basic properties of the three reasoning methods, some advantages can be found in the proposed scheme. An example of weld forming prediction is provided to illustrate the application of the algorithm.