基于多种群遗传算法与神经网络的激光-电弧复合焊接焊缝形貌预测

发布时间：2018-06-29 17:15

本文选题：激光-电弧复合焊接 + 神经网络　；参考：《应用激光》2015年06期

【摘要】：在激光-电弧复合焊接试验中,利用神经网络来模拟预测焊缝形貌时,由于众多的焊接工艺参数以及模型输出参数使得预测结果和试验真实形貌误差较大,而且模型对焊接参数变化不灵敏。因此,引入多种群遗传算法对神经网络进行初始权值和阈值逐代优化,可以避免未成熟的收敛问题,使得优化后的模型对焊缝形貌预测具有较高精度。同时,将模型的输入参数和输出参数转换,得到利用熔深、熔宽、余高三个形貌尺寸参数来预测焊接工艺参数的网络模型。综合上述模型得到一套由理想熔深、熔宽、余高的焊缝形貌尺寸对应的焊接工艺参数,进而模拟出具体的焊缝形貌曲线图的系统。通过优化模型预测的形貌和实际形貌对比,预测误差在5%以内,参数转化后,优化的预测模型误差在3%左右。结合两个预测模型的预测系统,最后得到的预测形貌和实际形貌误差在10%左右。此系统可以避免大量的试验,大大缩短复合焊接研究周期,对焊接工艺参数优化具有重要意义。
[Abstract]:In the laser-arc composite welding test, when using neural network to simulate and predict the weld morphology, because of the numerous welding process parameters and model output parameters, the error between the predicted results and the true shape of the test is large. Moreover, the model is insensitive to the variation of welding parameters. Therefore, introducing multi-population genetic algorithm to optimize the initial weights and thresholds of neural networks from generation to generation can avoid the premature convergence problem and make the optimized model have a high precision for the prediction of weld morphology. At the same time, by converting the input parameters and output parameters of the model, a network model is developed to predict the welding process parameters by using the parameters of depth, width and residual height. Based on the above model, a set of welding process parameters corresponding to the size of ideal weld depth, weld width and residual height were obtained, and the specific weld profile curves were simulated. The prediction error is less than 5% and the error of the optimized prediction model is about 3% after parameter transformation. Combined with the prediction system of the two prediction models, the error of the predicted morphology and the actual morphology is about 10%. This system can avoid a large number of tests and greatly shorten the research period of composite welding, which is of great significance to the optimization of welding process parameters.
【作者单位】：长春理工大学机电工程学院;
【基金】：国家青年科学基金资助项目(项目编号:51305044)
【分类号】：TP18;TG456.9

【参考文献】