基于人工神经网络的激光点焊焊点形态预测

建立适用于激光点焊焊点形态预测的人工神经网络模型，以点焊过程中的三个主要工艺参数(激光功率、点焊时间和离焦量)作为模型输入，输出为焊点表面、熔合面、背面直径以及熔深和横截面面积五个焊点形态参数。在此基础上，建立焊点形态模型，模型输入为神经网络的预测结果，输出为焊点形态。所建立的神经网络预测模型和焊点形态模型结合之后，可以实现激光点焊焊点的形态预测。网络测试结果显示实际值与网络预测值之间的RMS误差为0.1左右，模型输出的预测焊点形态与实际焊点形态之间较为吻合。根据模型的仿真结果，进一步研究点焊参数对焊点尺寸和形态的影响规律。结果表明未熔透焊点形态为Y形，而熔透焊点则存在多种形态，形态之间的转变主要受激光功率的影响。

Prediction of Laser Spot Weld Shape by Using Artificial Neural Network

An artificial neural network (ANN) and a weld shape model are developed for the analysis and simulation of the correlation between the laser spot welding (LSP) parameters and weld shapes. The input parameters of the ANN consist of laser power, welding time and defocusing. The output is weld shape parameters namely: surface diameter, bottom diameter, nugget diameter, penetration depth and cross section area. The shape parameters predicted by the ANN are converted into a weld profile by the weld shape model. The combined influence of laser power and welding time on the weld shape is simulated. A comparison is made between measured and calculated data. The calculated results are in good agreement with measured data.