摘要
对于误差逆传播 (BP)神经网络标准及改进型算法中的神经元联接权更新机制进行了分析。作为实际应用 ,基于MATLAB神经网络工具箱函数 ,研究了两个例子 ,一个是不同合金种类及不同复杂程度的型腔结构下浇注温度的选择 ;另一个是压铸工艺的缺陷分析。采用的算法分别为恒定学习率 附加动量项改进型BP算法和自适应学习率 附加动量项改进型BP算法。通过在实际模拟中比较这两种算法的训练效率 ,可得到结论 :自适应学习率 附加动量项算法是精确模拟压铸工艺映射问题的较为理想的方法。基于这一方法 。
The connective weight updating mechanism of Back error Propagation (BP) neural network, especially its improved update algorithm was analysed. With Matlab neural network toolbox, the BP algorithm was adopted to simulate selection criteria of pouring temperature under different alloy category and die cavity complexity, and die casting defects pattern recognition, based on the method of constant learning rate plus momentum, and the method of adaptive learning rate plus momentum respectively. By comparing the training efficiency of the two algorithms, it is concluded that the BP neural network algorithm based on adaptive learning rate is more efficient for simulating processing mapping relations in die casting. The results of simulation testing for the networks were provided.
出处
《机械科学与技术》
CSCD
北大核心
2004年第7期816-819,共4页
Mechanical Science and Technology for Aerospace Engineering
基金
国家留学基金项目 (2 185 2 0 3 5 )
贵州省教委科研基金项目 (19980 17)资助
关键词
压铸
神经网络
浇注温度
缺陷分析
Die casting
Neural network
Pouring temperature
Defects pattern recognition
