基于机器学习的铝合金脉冲激光焊接工艺及质量预测研究

Quality Prediction of Pulse Laser Welding of Aluminum Alloy Based on Machine Learning

铝合金脉冲激光焊接能够精准调节激光能量输入,广泛应用于动力电池与新能源汽车等精密加工领域。然而,铝合金自身高导热率和高反射率等固有属性,以及与高能量激光剧烈的相互耦合作用,对工艺参数优化和焊接质量控制带来挑战。以2 mm厚1060铝合金作为研究对象,主要分析了不同脉冲激光工艺参数(峰值功率、脉冲频率和焊接速度)对焊缝成形的影响规律;以工艺参数为多维输入变量,进一步构建了基于BP神经网络的熔池尺寸预测模型。结果表明:不同工艺参数均对焊缝熔深和熔宽有直接影响,需要确定一个合适的工艺窗口;同时构建模型的平均预测误差在10%以内,具有较高的预测精度。为铝合金脉冲激光焊接质量预测及工艺优化提供了可靠的实验和指导依据。

Pulsed laser welding of aluminum alloy is widely used in precision machining fields such as power batteries and new energy vehicles because it can accurately adjust the laser energy input.However,the inherent properties of aluminum alloy such as high thermal conductivity and high reflectivity,as well as the intense interaction with high energy laser,bring challenges to the optimization of process parameters and welding quality assurance.In this paper,the influences of different pulse laser parameters(peak power,pulse frequency and welding speed)on weld forming were analyzed.Taking the process parameters as multi-dimensional input features,the prediction model of melting width and depth based on BP neural network was further constructed.The results show that different process parameters have direct influences on the weld depth and width,and it is necessary to determine a suitable process window.At the same time,the average prediction error of the constructed model is less than 10%,which has a high prediction accuracy.The research provides a reliable experimental and guiding basis for the quality prediction and process optimization of pulsed laser welding of aluminum alloy.