Artificial neural network application to study quantitative relationship between silicide and fracture toughness of Nb-Si alloys

In this paper, a forward and backward feed propagation artificial neural network (BP ANN) was developed to research the quantitative relationship between silicide and fracture toughness of Nb-Si alloys. The alloys were produced by directional solidification and heat treatment. The toughness was measured by a three-point bending method used for ANN output. Five characteristic factors used for ANN input were abstracted and measured. The sequence of factors is silicide volume fraction > gamma-Nb5Si3 > silicide shape > silicide size > silicide continuity by sensitivity analysis. As a result of this study, the ANN model was found to be successful for predicting toughness with high accuracy and good generalization ability within the range of 9.2-26.1 MPa m(1/2). The quantitative formulas of silicide feature parameters and fracture toughness were established by transfer function, weight matrix and threshold of ANN model. The effect of each parameter and interact influence of two parameters on the fracture toughness were studied, and the technological parameters of the alloy were optimized by artificial neural network model.