基于ELM的齿轮钢淬透性预测模型

Predicting model of gear steel hardenability based on extreme learning machine

利用极限学习机(ELM)研究了20Cr齿轮钢端淬硬度曲线和化学成分的预测,并将其预测结果与传统预测模型进行比较。结果表明,ELM可以根据齿轮钢的化学成分预测其淬透性,计算精度明显优于传统线性拟合及神经网络模型,同时ELM也能通过齿轮钢淬透性硬度曲线反测化学成分,元素含量误差在5%以内。

Prediction of end-quenching hardness curve and chemical composition of 20 Cr gear steel was studied by using extreme learning machine(ELM),and the prediction results were compared with the traditional prediction models.The results show that the ELM not only can predict the hardenability of the gear steel according to the chemical composition with calculation accuracy significantly higher than that of the traditional linear fitting and neural network models,but also can be used to backward predict the chemical composition from the hardenability curve with the element content error within 5%.