摘要
将人工神经网络理论及Back propagation(BP)算法应用于双层辉光等离子渗金属工艺的研究,并针对BP神经网络收敛速度慢、易陷入局部极小的缺点,提出一种新的动态退火算法优化网络的训练,进而建立了双层辉光等离子渗金属工艺参数与渗层元素总质量分数、渗层厚度和表面硬度之间的数学模型,最后将模拟预测结果与实验数据进行比较和误差分析,证明该模型具有较高的预测精度。
The Artificial Neural Network and Back propagation (BP) algorithm are applied in the double glow plasma surface alloying. To avoid long training time and local minimum points during training BP network, this paper provided a new Dynamic Annealing algorithm that can be used to optimize the training of neural network. Then, the mathematic model was built based on the relationship between technological parameters of double glow plasma surface alloying processing and the composition, gross mass fraction of element, thickness of surface alloying layer and surface hardness. Finally, it is proved that this model has a higher predictive accuracy through error analysis and comparing the data between theoretical model and experiment.
出处
《材料科学与工程学报》
CAS
CSCD
北大核心
2007年第3期426-429,共4页
Journal of Materials Science and Engineering
基金
江苏省自然科学基金资助项目(BK2005128)
关键词
双层辉光
数学模型
人工神经网络
动态退火算法
double glow
mathematic model
Artificial Neural Network
Dynamic Annealing algorithm