摘要
以合金成分、浇注温度、型腔真空压力、铸造压力和模具温度作为输入层参数,力学性能(抗拉强度)和耐腐蚀性能(腐蚀电位)作为输出层参数,构建了3×15×1三层拓扑的遗传算法BP神经网络优化模型,并对该模型进行了预测和验证。结果表明,模型输出的抗拉强度平均相对训练误差为3.46%,平均预测误差值为3.40%;腐蚀电位平均相对训练误差为3.78%,平均预测误差值为3.59%,模型预测能力强、精度高。
Taking alloy composition, pouring temperature, cavity vacuum pressure, casting pressure and die temperature as input layer parameters, and mechanical properties(tensile strength) and corrosion resistance(corrosion potential) as output layer parameters, a genetic algorithm BP neural network optimization model with 3×15×1 three-layer topology was constructed, and the model was predicted and validated. The results show that the average relative training error of tensile strength outputted by the model is 3.46%,and the average forecast error is 3.40%. The average relative training error of corrosion potential is 3.78%, and the average prediction error is 3.59%. The model has strong prediction ability and high accuracy.
作者
张锴
孙冬
ZHANG Kai;SUN Dong(Jilin Engineering Vocational College,Siping 136001,China;Experimental Management Center,Henan Institute of Technology,Xinxiang 453000,China)
出处
《热加工工艺》
北大核心
2020年第21期67-70,共4页
Hot Working Technology
基金
吉林省高等教育学会高教科研课题(C170996)。
关键词
BP神经网络
遗传算法
镁合金建筑模板
力学性能
耐腐蚀性能
BP neural network
genetic algorithm
magnesium alloy building formwork
mechanical properties
corrosion resistance
