Predicting mechanical properties of elastomers with neural networks

Despite the existence of a solid theoretical basis interrelating various mechanical properties of elastomers, the complexity of these materials and strong dependence of characteristic material parameters on deformational and temperature conditions cause insuperable difficulties in establishing accurate relations between the crosslinking properties of elastomeric compounds and crosslinked elastomers in practice. Since knowledge of such presumably nonlinear relations would be valuable for several reasons, this work attempts to uncover these relations using methods of soft computing, in particular neural networks. The resulting relations obtained by neural network analysis have proved to be incontestably good and completely in accordance with expectations, thus contributing to proficiency in dealing with elastomeric materials, as well as curtailing possibilities of testing redundancy.