摘要
对齿轮钢淬透性预报的研究现状进行了综述,结合人工神经网络理论在预测模型方面的应用,运用BP神经网络对转炉连续生产的100炉实际数据进行齿轮钢淬透性回归分析。结果表明,实测值与预测数据线性相关系数均在0.93以上,端淬控制模型具有高的准确性,可实现窄淬透性的控制。采用优化后的淬透性预报模型成功地进行了20CrMnTiH子钢号的开发,实现了窄成分的控制。
The paper describes the research situation of gear steel hardenability prediction, and makes the regression analysis of gear steel hardenability of the real convertor data of 100 heats in successive pro- duction using the feed-forward back propagation （BP） ANN, by combining with the artificial neural net- work （ANN） applied in the prediction model. The results show that the linearly dependent coefficient be- tween measured value and predicted data equals at 0.93 upwards, the end hardenability model has high accuracy and it can be applied in control narrow quenching degree predicting with low error. By applying the optimized quenching degree prediction model, the composition of 20CrMnTiH is narrow and the 20CrMnTiH subclass was successful developed.
出处
《冶金丛刊》
2013年第4期14-15,31,共3页
Metallurgical Collections
关键词
人工神经网络
成分
淬透性
artificial neural network
composition
quenching degree
