摘要
通过一种镁锂合金在Gleeble3500热模拟机进行的热压缩实验数据进行训练,模型能较准确地预测该材料的流变应力,误差低于5%.改进算法避免了标准BP网络易陷入局部最小以及收敛速度慢的缺点,得到了更高的精度以及训练速度.预报模型准确度及可靠性高,具有工程应用价值.
The flow stress model during Mg-Li Alloy hot deformation was built by improving BP ANN. The data of Mg-Li Alloy for ANN training was test by Gleeble3500 during hotcompression. The ANN can predict the material flow stress exactly and the error less than 5%. The improving BP algorithm freed the defect of local minimum value and lower speed of training when using BP algorithm. The predicted stress-strain curves are in good agreement with the experimental results.
出处
《数学的实践与认识》
CSCD
北大核心
2010年第9期87-92,共6页
Mathematics in Practice and Theory
