用人工神经网络研究钢的硬度的影响因素

Influencing Factors of Steel′s Hardness by Artificial Neural Network

用人工神经网络研究了化学成分及热处理工艺参数对低碳低合金钢的硬度的影响。首先设计了RBF型人工神经网络模型,用"舍一法"改进了模型,使其具有较好的预测性能。然后,用神经网络研究了化学成分和冷速对低碳低合金钢的硬度的定量影响。结果表明,碳的质量分数为0.11%～0.15%时,硬度随碳含量的增加而增大;硅的质量分数为0.24%～0.38%、锰的质量分数为0.94%～1.02%时,硬度值基本不变;铬的质量分数为0～0.6%时,硬度值呈增加趋势;镍的质量分数为0～0.04%时,硬度值基本不变;钼的质量分数为0～0.2%时,硬度值从HV 288降至HV 282;硼的质量分数为1%～2%时,硬度随含量增加而升高;钛、铌、钒的总质量分数为0.06%～0.14%时,硬度值基本不变;冷速从10℃/m增加至170℃/m,硬度值从HV 290增至HV 420。

The artificial neural network was used to research the influence of chemical composition and heat retreatment parameters on the hardness of the steel.Firstly,RBF artificial neural network was established to analyze the relationship of the chemical compositions-cooling rate-hardness,using the method of ＇eave-one-out＇ to practice the model to achieve good prediction performance.Then,quantitative influence of composition and cooling rate on the hardness of low carbon and low alloy steels was investigated using the neural network.The result shows that,the hardness increases with the carbon content from 0.11% to 0.15%;while silicon content is 0.24%-0.38% and manganese content is 0.94%-1.02%,the hardness value basically unchanged;The hardness value has the increasing trend with chrome content is in 0-0.6%;While nickel content is 0-0.04%,hardness value basically unchanged;molybdenum content is 0-0.2%,hardness value falls from HV 288 to HV 282;With boron content rises from 1% to 2%,hardness increases;The total content of vanadium,titanium,niobium changes between 0.06%-0.14%,the hardness value basically unchanged;Cooling rate increases from 10 to 170 ℃/m,hardness value increases from HV 290 to HV 420.