【作者】 曲周德； 张士宏； 王忠堂； 李殿中；

【Author】 Qu Zhoude Zhang Shihong Wang Zhongtang Li Dianzhong(Institute of Metal Research, Chinese Academy of Sciences, Shenyang, 110016) (Material Science and Engineering School, Taiyuan Heavy Machinary Institute,Taiyuan, 030024)

【机构】 中国科学院金属研究所；

【摘要】 本文对RBF(Radial Basis Function Network),MLP(Multi-Layer PerceptronNetwork),LM(Levenberg-Marquardt算法)等神经网络工作原理作了介绍,对微合金钢的力学性能、各相的分数和铁素体的晶粒尺寸之间的关系进行了研究,认为可以利用人工神经元方法分析微合金钢性能、组织成分及铁素体晶粒尺寸之间关系,利用试验数据通过多次计算机试验,并比较神经网络训练实验数据,预测结果的精度,比较结果表明,LM神经网络方法是用于组织性能预报的有效方法之一。更多还原

【Abstract】 Artificial Neural Network method such as RBF ( Radial Basis Function) Network, MLP ( Multi-Layer Perceptron) Network, LM ( Levenberg-?Marquardt) were introduced. The relationship between mechanical properties and volume of each phase and ferrite grain size was analyzed. Based on the experimental data, some computer tests were carried . The precision of prediction and validity of training was analyzed. It is concluded that LM is an effective method for the prediction of properties.更多还原