Laser welding is a thermal processing technology that utilizes a high-energy laser beam as a heat source. It boasts several advantages such as low heat input, narrow and high-quality weld seams, fast welding speeds, and high weld joint strength. These attributes have led to its widespread application in industries such as automotive components and large-scale machinery manufacturing. However, the quality of laser welding is influenced by various factors including process disturbances, material properties, and workpiece preparation, which can result in unstable welding quality and defects.

To address these challenges, researchers have developed various quality inspection technologies. Common techniques include X-ray inspection, acoustic emission testing, and plasma radiation and spatter detection. These methods employ different sensors and approaches to monitor physical signals during the welding process, such as melt pool shape, plasma radiation, and acoustic waves. By analyzing changes and characteristics in these signals, defects like porosity, cracks, and slag inclusions that may occur during welding can be detected promptly. This allows for timely adjustments to be made to ensure welding quality.

Studies indicate that these inspection technologies not only facilitate real-time monitoring of the welding process but also provide data support for optimizing welding parameters, thereby enhancing product quality and production efficiency. Looking ahead, with further advancements in sensor technology and data processing algorithms, laser welding quality inspection techniques are expected to become more intelligent and automated. This will contribute to greater benefits and development opportunities in industrial manufacturing.

In laser welding, the interaction between the laser beam and the material generates various physical effects that can be utilized for detection purposes. The information available for detection includes:

1. Radiation from the melt pool material: This refers to the thermal radiation emitted by the material in the melt pool during the welding process.

2. Electromagnetic radiation from plasma: Plasma generated during laser welding emits electromagnetic radiation across a spectrum including ultraviolet (UV), visible light, and near-infrared wavelengths.

3. Laser radiation reflected by the material: The laser beam directed onto the material reflects and can be detected to monitor the welding process.

4. Mechanical waves and acoustic waves: These are generated by material evaporation, thermal expansion, and oscillations in the gas environment during welding.

5. Charge accumulation on conductive materials in the plasma region: This accumulation of charge results in currents due to the movement of charged particles.

These detection methods utilize various sensors and detectors to monitor these physical signals during laser welding. They play a crucial role in real-time monitoring and quality control by detecting defects such as cracks, pores, and spatters, thereby ensuring the stability and quality of the welding process.

Quality inspection techniques during laser welding processes mainly involve the following methods and tools:

1. Photoelectric Diode/CCD Camera Inspection:

• Principle: Detect radiation signals generated when the laser interacts with the material, such as melt pool radiation and reflected laser radiation.

• Application: Monitor melt pool shape, size, and stability; analyze weld seam formation and quality issues using image processing technology.

2. Acoustic Emission Testing:

• Principle: Detect acoustic or ultrasonic signals caused by plasma movement during welding.

• Application: Monitor defects like porosity and cracks in real-time, providing early defect detection and warning.

3. Charge Collection Device:

• Principle: Detect plasma charge accumulation during welding.

• Application: Analyze plasma movement to monitor welding stability and melt pool behavior.

4. Spectral Analysis Technology:

• Principle: Analyze spectral characteristics generated during laser-material interaction.

• Application: Diagnose weld quality and detect defects by analyzing spectral signals of different wavelengths, providing information on material composition and temperature.

5. X-ray Inspection Technology:

• Principle: Use X-rays to penetrate materials and detect attenuation levels for internal structure and defect detection.

• Application: Detect internal defects in weld joints such as porosity and slag inclusion, providing high-resolution image data.

6. Machine Vision Technology:

• Principle: Analyze welding process video images using digital image processing techniques.

• Application: Real-time monitoring of melt pool morphology, weld seam formation process, and quality assessment, offering quantitative quality evaluation and defect detection.

These techniques are typically combined to collect, analyze, and process multiple signals and data generated during laser welding processes. They ensure real-time monitoring and control of welding quality, ensuring process stability and weld seam quality. With ongoing technological advancements and deeper application, laser welding quality inspection techniques continue to improve and optimize.

In practical production, several common defects in laser welding include porosity, slag inclusion, and cracks. Among these, accidents caused by cracks are the most common in post-weld inspection and engineering applications. Cracks mainly occur because welding stresses break the original covalent bonds between metals, and their small size often makes them difficult to detect with the naked eye. Porosity forms when air fails to escape during the solidification of the weld pool, creating voids inside the metal, which significantly reduces the toughness and fatigue strength of the weld joint. Slag inclusion occurs due to the presence of residual slag or metal debris in the weld pool, leading to irregular defects. Welding too slowly or with excessive current can result in weld beads, while welding too quickly or with insufficient current can create "undercuts," known as "bite edges." Common types of welding defects are shown in Figure 2.

图2 常见焊接缺陷类型^[22]

1. Laser Welding Feature Detection 3.1 X-ray Inspection Technology

X-ray inspection technology is a commonly used non-destructive testing method for assessing the quality of laser welding. It utilizes the penetrating properties of X-rays to examine internal defects and structural characteristics within the welding area. Specifically, X-rays can pass through metal weld seams and transmit images to detectors, which can then analyze these images to detect possible defects such as porosity, cracks, or slag inclusions.

The advantage of X-ray inspection technology lies in its ability to penetrate deep into materials and detect small and subsurface defects that may not be observable through visual or other surface inspection methods. By adjusting the energy and angle of X-rays, high-precision detection can be achieved for different materials and welding conditions. However, X-ray inspection also poses challenges such as radiation safety and high equipment costs, requiring strict control of radiation doses during operation to ensure the safety of operators and the environment.

In summary, X-ray inspection technology plays a crucial role in laser welding quality inspection, providing a reliable non-destructive assessment method to ensure the quality of weld joints.

3.2 Acoustic Emission Testing Technology

Acoustic emission testing technology, as a type of non-destructive testing, involves signals that propagate along the surface of the weld in the ultrasonic and audible sound wave ranges [39]. According to references [40-42], audible sound signals are caused by pressure fluctuations during the ejection of high-temperature plasma from pores. Research indicates that acoustic emission signals are associated with various factors such as post-weld cracks, welding stress relief, impact, and friction [39] [43]. Mengyu Chai [44] and Zhu Yang [45] explored the correlation between welding stability, weld defects, and characteristics of acoustic emission sources in experiments with laser welding of 304 stainless steel. Reference [46], focusing on MAG high-strength steel, extracted acoustic emission signals under conditions with and without welding defects. The conclusion suggests that acoustic emission signals can serve as reliable indicators for monitoring welding quality in shipbuilding thick plate welding. On the other hand, the nonlinear mapping relationship between acoustic emission signals and emission sources often poses challenges in establishing mathematical models [39]. This has led to widespread application of Artificial Neural Networks (ANN) in defect acoustic emission monitoring [47], as highlighted in references [48-50]. Through analysis of molten pool dynamics during welding, a BP neural network was constructed using collected feature parameters as inputs. The model effectively predicts and alerts to post-weld crack formation, and reference [50] also utilized the BP neural network system to predict the weldability of 304 stainless steel plates. Moreover, BP neural networks play a guiding role in localizing crack emission sources, suggesting optimal sensor deployment for precise defect location [51] [52]. Additionally, references [53-54] introduce filtering and frequency decomposition of arc sound signals and electric signals based on human auditory models for MIG welding, achieving recognition of three states: short-circuit, droplet transition, and jet transition, validating the model's interference resistance and proposing new online monitoring solutions for MIG welding quality.

3.3 Plasma Plume and Spatter Detection

During laser welding, rapid oscillations occur in the liquid and gas phases of the weld pool [10], transforming the welding process into a highly dynamic non-steady state [55]. Photogenerated plasmas in laser welding refract, absorb, and scatter incident laser energy [56-59], with studies in reference [15] demonstrating that intense plasma fluctuations lead to corresponding fluctuations in laser energy input, ultimately resulting in welding defects. Wang Nian, Shen Hua [60] addressed the need for large-caliber observation equipment for laser welding plasma radiation, proposing an optical imaging-based spectral radiation transfer model to accurately invert photogenerated plasma radiation fields. Reference [61] suggests a direct correlation between laser welding plasma charge voltage and electron temperature, identifying phases of plasma charge voltage variation (excitation, maintenance, decay), providing theoretical support for plasma monitoring technology in laser welding processes.

At Southwest Jiaotong University, Ma Yaorui, Cai Chuang [55] designed a laser-MIG hybrid welding monitoring system in LabVIEW environment, utilizing binarization, morphological filters, HUBER linear fitting, and branch removal algorithms for workpiece identification, successfully determining laser transmission distances in photogenerated plasmas. Zhao Shengbin, Yang Lijun [62] utilized passive electrical detection devices to collect plasma electrical signals, concluding that the plasma signal of deep penetration welding exhibits an initial peak phase, followed by oscillatory decay over time to a stable state. In contrast, heat conduction welding exhibits stable oscillation throughout the welding process without an initial peak phase. Furthermore, reference [62] analyzed the mechanism of internal plasma current generation, determining that changes in laser plasma signal characteristics are jointly determined by plasma and sheath effects, proposing a new identification method for initial stage welding modes in laser welding. Huang Ruisheng, Zou Jipeng [63] compared surface flow characteristics of molten pools and plasma plume fluctuation characteristics between single laser welding and laser scanning welding conditions, concluding that oscillating laser beam technology enhances plasma stability relative to single laser beam technology, thereby improving pore stability and reducing defect formation.