论文部分内容阅读
针对LD封装激光焊接后变形产生的耦合偏差,本文首先设计了不同参数对焊后偏移(PWS)影响程度的实验,通过参数间排列组合的变化观测其对PWS的影响程度;进而设计出神经网络焊后偏移预测模型,并将预偏移大小、焊点位置、激光脉冲能量以及焊接前初始位置等可以实时调节的焊接参数作为神经网络输入,模型的PWS预测值与实际值吻合效果良好。实验结果表明,利用神经网络预测PWS的方法是可行的,并且在实际焊接过程中,可以借助神经网络预测模型选择最佳的焊接初始条件使PWS降至最低,大大提高了LD的封装效率和产品可靠性。
Aiming at the coupling deviation caused by deformation after laser welding of LD package, this paper first designs experiments on the effect of different parameters on post-weld deflection (PWS), observes its influence on PWS by the change of the arrangement and combination of parameters, and then designs the neural The prediction model of the post-welding offset is put forward, and the welding parameter which can be adjusted in real time, such as the pre-deflection size, the position of the welding spot, the laser pulse energy and the initial position before welding, is input as a neural network, and the PWS predicted value is in good agreement with the actual value . The experimental results show that it is feasible to predict the PWS by neural network. In the actual welding process, the optimal welding initial conditions can be selected by the neural network prediction model to minimize the PWS and greatly improve the packaging efficiency of LD and the product reliability.