Some conditions to use the fuzzy logic model for approximate reasoning

The goal of this paper is to present some conditions to be imposed to the fuzzy model so that it works and gives satisfactory results in approximate reasoning applications. The main problem of this model arises when several conclusions have to be combine because the result could be a non-normal fuzzy set with a very irregular shape or it could be an empty set. The subnormal and irregular fuzzy set is very hard to approximate using linguistic terms or numerical values. These problems can be overcome imposing some conditions to the model, in particular to the fuzzy partitions. Under these conditions, the output is a normal fuzzy set and the empty conclusion is avoided. Besides, the final result can be easier to approximate with linguistic terms. Also, the quantification of the output allows a more objective way to do this approximation. These results are compared with those obtained with a new approach based on the defuzzification of the output.