Artificial neural network prediction of heat-treatment hardness and abrasive wear resistance of High-Vanadium High-Speed Steel (HVHSS)

Abstract

The hardness and abrasive wear resistance were measured after High-Vanadium High-Speed Steel (HVHSS) were quenched at 900 °C–1100 °C, and then tempered at 250 °C–600 °C. Via one-hidden-layer and two-hidden-layer Back-Propagation (BP) neural networks, the non-linear relationships of hardness (H) and abrasive wear resistance (e) vs. quenching temperature and tempering temperature (T1, T2) were established, respectively, on the base of the experimental data. The results show that the well-trained two-hidden-layer networks have rather smaller training errors and much better generalization performance compared with well-trained one-hidden-layer neural networks, and can precisely predict hardness and abrasive wear resistance according to quenching and tempering temperatures. The prediction values have sufficiently mined the basic domain knowledge of heat treatment process of HVHSS. Therefore, a new way of predicting hardness and wear resistance according to heat treatment technique was provided by the authors.