engleski

Mechanical Properties of Ductile Cast Iron Determined by Neural Networks

This paper presents the results of application of artificial neural networks in determination of mechanical properties of ductile cast iron. All data were collected in a Croatian foundry for 147 melts. Error back-propagation training algorithm was applied to train the multilayer networks. The optimal size of the hidden neuron layer was determined through the analysis of error parameters. The optimal slope of the tangent sigmoid activation function was investigated. Neural networks were modeled to predict each mechanical property, namely: tensile strength, yield strength, elongation, and hardness. Input parameters were 13 weight contents of chemical elements in the melt. A comprehensive analysis of errors in predicting mechanical properties of ductile cast iron was made.

Determination; Ductile Cast Iron; Mechanical Properties; Neural Networks; Prediction

nije evidentirano