Flow Stress Prediction of Hot Compressed Tool Steel by CAE NN and Hyperbolic-Sine Equation

摘要

Hot compression experiments are carried out on steel workpieces by means of Gleeble 1500 thermo mechanical simulator in wide range of temperatures 800 deg C -1200 deg C with strain rates 0,1 s~(-1) 1,0 S~(-1) and 8,0 s~(-1) and true strains of 0,0 to 0,5. Hot flow curves were estimated by means of the CAE neural networks. The methods of constant smoothness parameter and non-constant (ellipsoidal) smoothness parameter were applied. The use of the latter proved more exact (up to 3,4 percent) and simpler if we compare it with the existing data for the flow curve prediction of tool steel by BP NN (up to 7 percent), as the proposed method yields better results. The activation energy and other parameters in hyperbolic-sine equation were calculated according to the method proposed by McQueen et al. and according to the method recently proposed by Kugler et al. The latter yields better results at predicting the maximum values of hot flow curves.