基于光谱诊断的钛合金激光焊接熔深在线检测

Online Detection of Laser Welding Penetration Depth of Titanium Alloy Based on Spectral Diagnosis

激光焊接是一种多物理场耦合的高速激光加工过程,对其熔深的精确在线检测一直是行业最为关注的关键难题.本文针对TC4钛合金激光焊接过程,搭建了光致等离子体光谱实时监测平台,分析了激光功率对等离子体热力学特性、熔深的影响规律,提出了一种熔深在线预测方法.研究结果表明,等离子体光谱数据和熔深的变化存在显著相关性.基于此,选择了6条线谱的强度特征用于熔深预测建模.并采用降维技术对光谱数据进行特征提取,基于提取的特征再次建立了熔深预测模型.基于数据降维变换后的光谱特征构建的熔深预测模型的均方误差为0.0194,小于基于原始光谱强度特征建立的神经网络模型.本文构建的高精度熔深预测模型为激光焊接熔深的实时反馈控制提供了理论依据.

Laser welding is a high-speed laser process that couples multiple physical fields.The accurate online detection of the penetration depth of laser welding has been a critical challenge.Consequently,this study focused on the laser welding process of TC4 titanium alloy and constructed a spectrum monitoring platform for laser-induced plasma.The influence of laser power on plasma thermodynamic characteristics and penetration depth was analyzed.Subsequently,a real-time prediction method for penetration depth was proposed.The results determined a significant correlation between plasma spectral data and penetration depth values.Based on this correlation,the intensity characteristics of six line spectra were selected to construct a real-time prediction model of penetration depth.Using dimensionality reduction technology to extract features from spectral data,another prediction model for penetration depth was established based on the extracted features.The mean square error of the penetration depth prediction model constructed using spectral features after dimensionality reduction was 0.0194,which was smaller than that of the penetration depth prediction model based on the original spectral intensity features.Thus,the high-precision penetration depth prediction model constructed in this study provides a theoretical basis for real-time feedback control of laser welding penetration depth.