Evolutionary design of generalized polynomial neural networks for modelling and prediction of explosive forming process

Abstract Some aspects of explosive forming process have been investigated experimentally and modelled using generalized GMDH-type (group method of data handling) neural networks. In this approach, genetic algorithm (GA) and singular value decomposition (SVD) are deployed simultaneously for optimal design of both connectivity configuration and the values of coefficients, respectively, involved in GMDH-type neural networks which are used for modelling of centre deflection, hoop strain and thickness strain of explosive forming process. In particular, the aim of such modelling is to show how these characteristics, namely, the centre deflection, the hoop strain and the thickness strain change with the variation of important parameters involved in the explosive forming of plates. In this way, a new encoding scheme is presented to genetically design the generalized GMDH-type neural networks in which the connectivity configuration in such networks is not limited to adjacent layers. Such generalization of network's topology provides optimal networks in terms of hidden layers and/or number of neurons so that a polynomial expression for dependent variable of the process can be achieved consequently. It is also demonstrated that singular value decomposition (SVD) can be effectively used to find the vector of coefficients of quadratic sub-expressions embodied in such GMDH-type networks.