基于BP神经网络的连续油管焊接热影响区性能预测  

Prediction of mechanical properties of welding HAZ of coiled tubing based on BP neural network

作　　者：李继红[1] 李琳[1] 毕宗岳[1,2] 张敏[1] 

机构地区：[1]西安理工大学材料科学与工程学院,陕西西安710048 [2]宝鸡石油钢管有限责任公司,陕西宝鸡721008

出　　处：《兵器材料科学与工程》2012年第3期15-18,共4页Ordnance Material Science and Engineering

基　　金：陕西省教育厅自然科学基金资助项目(00k904);陕西省重点学科建设专项资金资助项目(00X901)

摘　　要：通过热模拟实验获得连续油管焊接热影响区最薄弱区的力学特性,采用BP神经网络对该区域在不同工艺下的力学性能进行仿真预测,着重研究不同训练函数对网络性能的影响。通过不同训练函数条件下网络性能的对比分析,最终选取最优的网络模型对连续油管焊接热影响区最薄弱区的力学特性进行预测,获得较好的预测结果。The mechanical properties of the weakest areas in welding HAZ of coiled tubing were obtained through thermal simulation experiment.The BP neural network was used to predict the process performance of the region.Based on the above testing or prediction,the influence of different training function on network performance was discussed in detail.The best network model applied to predict the mechanical properties of the weakest areas in welding HAZ of coiled tubing could be chosen through comparing and analyzing the network performance under different training function,so the preferable prediction results were obtained.

关 键 词：连续油管 热影响区 BP神经网络 力学特性 

分 类 号：TG445[金属学及工艺—焊接]