摘要
根据电化学极化法测定的10种热脆化程度不同的30Cr2MoV转子合金在不同钼酸钠电解液温度下的二次峰电流密度数据和合金的化学成分J参数数据,采用BP神经网络建立了转子合金脆性转变温度与二次峰电流密度、电解液温度、J参数之间的映射模型。通过对两种新30Cr2MoV转子合金的脆性转变温度进行预测,并将测算结果与线性回归方法得到的结果进行比较,结果表明网络训练误差和检验误差在±20℃以内,所建网络预测模型能较准确的预测新转子合金的脆性转变温度,BP神经网络用于转子合金的脆性转变温度的预测是有效、可行的。
Based on the data of the second peak current density （Ip） of ten types of 30Cr2MoV rotor steels with different fracture appearance transition temperature （FATT50） and the data of rotor alloying elements, a back-propagation （BP） neural network model is constructed and that the non-linear relationship between FATT50 and Ip, temperature of electrolyte, J-factor is established.The fracture appearance transition temperature of two different types of rotor alloy can be predicted by means of the trained neural network from the testing data, the train error and check error is with the scatter of ±20℃. The results show that, for the temper embrittlement of rotor alloy prediction, the prediction model based on BP artificial neural network is feasible and effective.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2005年第23期110-113,共4页
Proceedings of the CSEE
基金
国家电力公司科技项目(SP11-2001-02-29)
关键词
热能动力工程
汽轮机转子
神经网络
热脆性
电化学极化法
Thermal power engineering
Steam turbine rotor
Neural network
Temper embrittlement
Electrochemical technique
