连续油管TIG焊接头最薄弱区工艺-性能神经网络预测模型  

The Neural Network Prediction Model of Process-property in the Weakest Area of Coiled Tubing TIG Welded Joint

作　　者：李琳[1] 李继红[1] 余晗[2,3] 赵鹏康[1] 毕宗岳[1,2,3] 张敏[1] 

机构地区：[1]西安理工大学材料科学与工程学院,西安710048 [2]国家石油天然气管材工程技术研究中心,陕西宝鸡721008 [3]宝鸡石油钢管有限责任公司钢管研究院,陕西宝鸡721008

出　　处：《焊管》2012年第1期5-7,12,共4页Welded Pipe and Tube

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：通过试验检测出连续油管TIG焊接头最薄弱区力学性能,采用BP神经网络对该区域工艺性能进行仿真预测,研究了不同训练函数对网络性能的影响,通过对比分析不同训练函数下网络性能,得出连续油管TIG焊接头最薄弱区线能量-冲击功预测模型,最终选取LM算法、SCG算法对网络进行训练,采用这两种算法建立起的线能量-冲击功模型精度较高,测试数据预测值与实测值平均相对误差分别为0.785%和0.34%,网络能够很好地对"未知"冲击功进行预测。The mechanical properties of the weakest areas in the joint of coiled tube welded by TIG was obtained according to the experiment. BP neural network was used to simulate and predict the process performance of the region, The influence on network performance was studied under different training function, Line Energy-Impact energy prediction model of the weakest areas in the joint of coiled tube welded by TIG was obtained by comparing the Network performance which received under different training function. LM algorithm and SCG algorithm was selected to train the network finally. Both the algorithms present higher precision of Line Energy-Impact energy prediction model. The average relative error of predicted and measured values of test data were 0.785% and 0.34% respectively. It was very well that the impact energy were predicted in the network.

关 键 词：连续油管 TIG焊 BP神经网络 冲击韧性 

分 类 号：TE973[石油与天然气工程—石油机械设备]