Comparison of two methods for the analysis of composite material

The subject of this paper is the neural network and linear regression as tools for the estimation of the component–property relationship of composite materials. Test data were used for the teaching of an ANN, verifying its output and computing a representation of the regression polynomial. Two kinds of expert systems have been built and tested for different sets of input and output characterizing the component and the properties of the material. Numerical experiments were conducted for real data given by a concrete factory. The linear regression method and the developed ANNs can be used for predicting material properties when the components are known as well as for the inverted problem. The expert systems were described and compared.