摘要
根据A356泵体铸件金属型低压铸造特点,结合生产实际,以A356泵体浇注工艺参数为研究对象,L16(45)型正交实验数据作为训练学习样本,与正交实验成分有关的前16个样本作为训练与检验,用BP神经网络进行预测和优化,结果表明神经网络优化后的模拟值最大误差很小,CPU占用时间仅为40s。人工神经网络与正交实验相结合,能大大节省时间和费用,降低CPU占用率,也证实了对A356泵体充型过程数值模拟的神经网络优化是可行的。
According to the features of pump body A356 low-pressure die casting and the actual conditions, taking the pump body A356 low-pressure die casting process parameters as the studied object, the L16(4^5)-orthogonal experimental data as the trained samples, the random 16 samples related to the orthogonal test were forecast and assessed in the artificial neutral network of BP. After the optimization, the maximum relative errors of simulation is slight, and its CPU time is 40 s. The combination between artificial neural network and orthogonal experiment greatly saves time and costs, lowers CPU occupancy rate, which confirmed that neuro-optimization of numerical simulation in solidification process of pump body A356 is feasible.
出处
《热加工工艺》
CSCD
北大核心
2009年第13期58-60,64,共4页
Hot Working Technology
关键词
低压铸造
数值模拟
BP神经网络
正交实验
优化
low-pressure die casting
numerical simulation
BP algorithm
orthogonal design
optimization