| 引用本文: | 韩力,王天琪,龙斌,王克宽.V型坡口多层多道焊焊道几何参数预测[J].材料科学与工艺,2023,31(5):38-44.DOI:10.11951/j.issn.1005-0299.20220173. |
| HAN Li,WANG Tianqi,LONG Bin,WANG Kekuan.Prediction of weld channel geometric parameters for V-bevel multi-layer multi-pass welds[J].Materials Science and Technology,2023,31(5):38-44.DOI:10.11951/j.issn.1005-0299.20220173. |
|
| 摘要: |
| 中厚板V型坡口多层多道焊焊接过程中,每一条焊道的几何参数都会影响最终焊缝成形质量。为了评估V型坡口多层多道焊缝成形质量,提出了基于麻雀搜索算法(SSA)优化的BP神经网络模型预测焊道几何参数。文中通过实验分析各焊接工艺参数对焊道成形几何尺寸的影响,确定了以焊接电流、焊接速度、熔池宽度作为模型的输入,将能表征焊道质量的焊道高度和焊道计算高度作为模型的输出。对优化前后BP神经网络预测模型的性能进行对比,结果表明,优化后模型预测结果的相对误差分别保持在±4%、±8%以内,模型的稳定性、准确率都有较大提升,证明了该方法可有效预测V型坡口焊接时的焊道几何参数。 |
| 关键词: 多层多道 焊缝成形 质量评估 焊道几何参数 麻雀搜索算法 神经网络 |
| DOI：10.11951/j.issn.1005-0299.20220173 |
| 分类号:TG409；TP183 |
| 文献标识码:A |
| 基金项目:天津市科技特派员项目(20YDTPJC00780)；天津市“项目+团队”重点培养专项资助 (XC202053)；天津市教委科研计划项目(2019KJ011). |
|
| Prediction of weld channel geometric parameters for V-bevel multi-layer multi-pass welds |
|
HAN Li1, WANG Tianqi1, LONG Bin2, WANG Kekuan2
|
|
(1.School of Mechanical Engineering, Tiangong University, Tianjin 300387, China; 2.CNPC Engineering Technology Research Co., Ltd., Tianjin 300451, China)
|
| Abstract: |
| The geometric parameters of each weld channel in the V-bevel multi-layer multi-pass welding process of medium-thick plate affect the final weld formation quality. For evaluating the V-bevel multi-layer multi-pass weld forming quality, a BP neural network model based on sparrow search algorithm (SSA) optimization was proposed to predict the weld channel geometric parameters. The influence of each welding process parameter on the weld channel forming geometry was experimentally analyzed, and the welding current, welding speed, and melt pool width were determined as the inputs of the model. The weld channel height and the calculated weld channel height that could characterize the weld channel quality were taken as the outputs of the model. The performance of the BP neural network prediction model before and after optimization was compared, and results showed that the relative error of the prediction results of the optimized model was kept within ±4% and ±8% respectively, and the stability and accuracy of the model were greatly improved, indicating that the method can effectively predict the geometric parameters of the weld channel during V-bevel welding. |
| Key words: multi-layer multi-pass weld forming quality assessment weld channel geometric parameters sparrow search algorithm neural network |
