Use of modular networks in the building of constant life diagrams

The objective of this paper is to assess the use of new network architectures in modelling and building of constant life diagrams, using a small number of S–N curves in the training set as a continuation of a previously published paper regarding an FNN (feed-forward neural network). Two architectural models of modular networks, one of them with a gating network very popular in the literature and the other one with a gating network that takes into account the type of load applied to the material, were used. As in the previous paper, data on fatigue to constant stress of a fibreglass composite was tested for 12 S–N curves different. Based on the results it was observed that the ANN having the gating network that takes into account the type of load applied is the one which showed more reliable results and that both modular ANN give better results than the FNN.