Type-1 and type-2 radial basis function neural networks Mandami system to evaluate quality features

Abstract

This paper presents a methodology that uses the central composite design and the radial basis function neural networks in type-1 or in interval type-2 model to generate a network that evaluates quality features in an industrial image processing. The methodology includes a couple of radial basis functions as Huygen’s tractrix and triangular membership functions as complementary contributions that have not been reported in literature as radial basis functions. The advantage of using this proposal is that the training is not required to get an accurate result, also the generation of the IT2 RBFNN fuzzy rule base for evaluating quality characteristics is simplified by using the central composite design method and statistical indicators extracted from the product specification data. Experimental results show an error reduction of 90% when the interval type-2 Mandami Radial basis function neural network was compared against its type-1 counterpart using the Gaussian membership functions onto a radial basis function network. On the other hand, the implementation of the Huygen’s tractrix, found a reduction error of 50% in comparison to the Gaussian function.