Laser deep penetration welding (LDPW) of high-reflectivity alloys presents significant challenges due to the complex interaction between laser and material. Traditional methods for monitoring weld penetration depth are often inefficient or lack real-time accuracy, leading to inconsistent welding quality. Optical coherence tomography (OCT) has emerged as a promising technique for non-destructive, real-time monitoring of weld depth. However, the effectiveness of OCT-based penetration depth monitoring heavily depends on algorithmic optimization. This study aims to develop an improved penetration depth monitoring algorithm tailored for OCT, enhancing its accuracy and robustness when applied to high-reflectivity alloy welding. By optimizing signal processing and feature extraction, this research seeks to advance real-time penetration depth measurement, contributing to more precise and efficient laser welding processes.

This study proposes an improved OCT weld penetration depth extraction algorithm, which optimizes the existing OCT measurement method through four key steps of “noise reduction?ghosting elimination?data fitting?depth correction.” First, background noise from the charge coupled device (CCD) camera is removed using brightness distribution analysis, where a threshold is set to retain only valid data points. Second, the local outlier factor (LOF) algorithm is employed to detect and eliminate optical ghosting effects by analyzing density variations, and the primary data region is identified using a depth-based segmentation approach. Third, a moving average filter is applied to the keyhole depth data to smooth measurement fluctuations and obtain a continuous depth curve. Finally, based on metallographic measurements, a multivariate regression model is established to analyze the influence of welding parameters (laser power, welding speed, defocus amount, etc.) on the correlation between keyhole depth and actual weld penetration depth. This model is used to correct OCT measurement errors. The optimized algorithm significantly improves the accuracy and reliability of OCT-based penetration depth monitoring, providing a more robust method for real-time welding depth measurement.

The proposed optimization algorithm effectively improves the accuracy of weld penetration depth measurement based on OCT. By integrating signal denoising, ghost image elimination, curve fitting, and penetration depth correction, the algorithm significantly enhances the reliability of depth estimation under various welding conditions. The results show that the optimized measurement data closely align with actual metallographic depths, with relatively small error, confirming its feasibility for online monitoring. Moreover, the algorithm performs well in both linear and oscillating welding processes, demonstrating adaptability to different materials and welding parameters. These verification results further validate the robustness and high precision of the algorithm, making it a viable solution for accurately measuring weld penetration depth in industrial applications.

This study addresses the challenge of low penetration depth measurement accuracy in laser deep penetration welding of highly reflective aluminum and copper alloys. An optimized OCT-based algorithm, incorporating signal denoising, ghosting elimination, data fitting, and depth correction, significantly enhances measurement precision by mitigating noise and ghosting interference. Validation results demonstrate that the optimized algorithm improves accuracy across different materials and welding parameters, maintaining error within 6%, making it a reliable approach for real-time weld quality monitoring and closed-loop control.