This paper presents an intelligent technology based method for analyzing and interpreting sensory data provided by multiple panels in evaluation of industrial products. In order to process the uncertainty existing in these sensory data, we first transform all sensory data on an unified optimal scale. Based on these normalized data sets, we compute the dissimilarities or distances between different panels and between different evaluation terms used by them, defined according to the degree of consistency of data variation. The obtained distances are then transformed into fuzzy numbers for physical interpretation. These fuzzy distances permit to characterize the evaluation behaviour of each panel and the quality of the evaluation terms used. Also, based on a Genetic Algorithm with punishment policy and the dissimilarity between terms, we develop a procedure for interpreting terms of one panel using those of another panel. This method has been applied to the fabric hand evaluation for a number of samples of knitted cotton in order to identify consumers' preference of different populations.
Linguistic classification: T-norms, fuzzy distances and fuzzy distinguishabilities
