Prediction of creep curve of HP40Nb steel using artificial neural network

摘要: Simulation of creep curves using data obtained from a limited number short-time tests is helpful for predicting the long-time life materials by extrapolation techniques. The present paper demonstrates application artificial neural network (ANN) prediction HP40Nb micro-alloyed steel. consists stress, temperature and time as input parameters strain output parameter. used are taken accelerated carried out at constant temperatures in range 650–1050 °C stresses 47–120 MPa. was trained three-layer feed-forward back-propagation network, having 15-neuron hidden layer, Levenberg–Marquardt optimization algorithm. After successful training, model subjected to several demonstrate consistent capability. 98% could be predicted within an error ±10% deviation experimental values. An additional experiment check capability confirms very good capability, with correlation coefficient 0.994, ANN modeling.