摘要
利用遗传算法的全局寻优能力对BP神经网络进行优化,改进BP神经网络易限于局部极小的缺陷。分析合金钢的CCT曲线特征,将CCT图分解为七个特征,并建立七个网络预测模型,实现合金钢的CCT曲线预测。结果表明,所建立的CCT曲线预测模型具有较理想的预测精度。
The local minimum problem of BP neural network was improved by using global optimizating ability of genetic algorithm to optimize BP neural network. CCT curve was decomposed into seven features, and seven network prediction models were established after analyzing the characteristics of the alloy steel. The results show that, the the established CCT curve prediction models have better prediction accuracy.
出处
《热加工工艺》
CSCD
北大核心
2014年第4期95-97,102,共4页
Hot Working Technology
