基于人工神经网络含稀土元素熔敷金属力学性能预测  

Prediction for mechanical properties of deposited metal containing rare earth elements based on artificial neural networks

作　　者：郭永环[1] 孟祥里 郭妍[2] 范希营[1] 张亮 

机构地区：[1]江苏师范大学机电工程学院,江苏徐州221116 [2]大庆油田管理局大庆钻探工程公司钻井生产技术服务二公司,黑龙江大庆163461

出　　处：《沈阳工业大学学报》2017年第3期269-274,共6页Journal of Shenyang University of Technology

基　　金：国家自然科学基金资助项目(51475220);江苏省产学研前瞻性联合研究项目(BY2016028-02);徐州市科技计划资助项目(KC15SM031)

摘　　要：为了提高焊条的力学性能并缩短焊条研发周期,在E4301型焊条药皮配方基础上加入了CeO_2和稀土元素La,并对焊条进行了力学性能试验.对试验数据进行分析后发现,加入适量的稀土元素可以改善焊条的力学性能.利用典型BP和RBF神经网络分别建立力学性能预测模型.将焊条中的CeO_2、La、Si、Mn含量与焊接速度作为预测模型的输入变量,将熔敷金属的抗拉强度、下屈服强度、断后伸长率与热影响区平均硬度作为输出变量.结果表明,将BP和RBF神经网络用于对含稀土焊条力学性能的预测是可行的,且RBF神经网络模型的预测精度和效率要高于BP神经网络模型.In order to enhance the mechanical properties of electrode and shorten the development cycle of electrode, the CeO2 and rare earth element （REE） La were added into the coating formula of E4301 electrode, and the mechanical properties of electrode were tested. Through analyzing the test data, it is found that the appropriate addition of REE can improve the mechanical properties of electrode. The prediction models for mechanical properties were established with BP and RBF neural networks, respectively. The contents of CeO2, La, Si and Mn in the electrode and the welding speed were taken as the input variables of prediction models. In addition, the tensile strength, lower yield strength, elongation and average hardness in the heat affected zone （HAZ） of deposited metal were taken as the output variables. The results show that it is feasible to use BP and RBF neural networks in predicting the mechanical properties of electrode containing REE. The prediction accuracy and efficiency of RBF neural network model are higher than those of BP neural network model.

关 键 词：La元素 焊接速度 BP神经网络 RBF神经网络 预测模型 熔敷金属 力学性能 焊条 

分 类 号：TG407[金属学及工艺—焊接]