基于人工神经网络及正交多项式回归分析的H13模具钢热处理工艺研究

Research of heat treatment process of H13 die steel based on artificial neural networks and orthogonal polynomial regression analysis

用BP人工神经网络、正交多项式回归分析及材料微观分析方法研究了热处理工艺对H13钢硬度的影响。结果表明,BP网络能根据淬火及回火温度精确预测H13钢热处理后的硬度。BP网络及正交多项式回归模型,均能用于研究淬火及回火温度对硬度的影响规律。模型研究结果表明,H13钢热处理时,淬火温度、第一次及第二次回火温度对硬度的影响显著,显著性水平为0.01。经980～1100℃淬火及540～620℃二次回火,H13钢的硬度在40～57 HRC范围;在给定的淬火温度下,随回火温度的提高硬度急剧降低,第二次回火温度比第一次回火温度对硬度的影响更大;在给定的回火温度下,淬火温度的增加使二次回火硬度增加约4 HRC;在1040℃淬火时,达到50～53 HRC硬度的回火温度区间最大;随淬火温度的提高,达到40～49 HRC硬度的回火温度区间减小。

Effect of heat treatment process on hardness of H13 steel was studied using a neural network with a back propagation algorithm （ BP network） , orthogonal polynomial regression and microstructural study methods. The results show that BP network can accurately predict the hardness of H13 steel according to quenching and tempering temperature. The trained BP network and hardness contour plot drawing from orthogonal polynomial regression equation were then applied to research the relation between hardness and quenching and tempering temperature. Quenching temperature, first and second tempering temperature have significant effect on hardness, the significance levels of which are 0.01. The hardness of H13 steel is in the range of 40-57 HRC, after quenched at 980-1100℃ and twice tempered in the range of 540-620 ℃. Hardness decreases with the increase of tempering temperature for certain quenching temperature. The second tempering has more significant influence on hardness than the first tempering. Tempering hardness increases about 4 HRC with quenching temperature increasing from 980 ℃ to 1100 ℃. When quenching temperature at 1040 ℃ , tempering temperature range achieves the maximum for tempering hardness in the range of 50-53 HRC, so those hardness can he obtained in wider tempering temperature range. Tempering temperature range decreases for the hardness in the range of 40-49 HRC, with quenching temperature increasing.