Application of generalized regression neural network optimized by fruit fly optimization algorithm for fracture toughness in a pearlitic steel

In this paper, a generalized regression neural network(GRNN) optimized by fruit fly optimization algorithm (FOA) was developed to improve the predictive accuracy of small number of samples. Then the quantitative relationship between alloying elements and fracture toughness was further investigated based on the developed model. The steels were produced by arc melting and heat treatment. The toughness was measured by a three-point bending method used for FOA-GRNN output. The characteristic factors used for input variables were alloying element content. The performance evaluation indices calculated for the improved models were compared with the results obtained by the GRNN model. As a result, the FOA-GRNN model was found to be successful for predicting toughness with high accuracy and good generalization ability for the studied steels. Through the sensitivity analysis, the results indicate that the sequence of influencing factors is Nb > C > Mn > S > Si > P > Cr > V. Then the interact effect of alloying elements on the fracture toughness were studied, and the composition parameters of the steel were optimized by the developed model.