Weld Defects Research Articles

Multi-model feature aggregation for classification of laser welding images with vision transformer

The laser welding technique is quite common in various manufacturing lines, including those for lithium-ion power batteries, due to its remarkable productivity, effectiveness, and flexibility. However, it has been observed that the consistency of the welding quality is not always optimal. This study investigates the adoption of laser welding in the manufacturing of lithium-ion batteries, such as their anode, cathode and safety vent, where consistent weld quality is crucial for battery performance and safety. Therefore, evaluating the laser-welding product is indispensable for industrial production and the public domain. Several techniques have been utilized for laser welding evaluation, both destructive and non-destructive. However, these methods are ineffective and too cumbersome to be adopted in mass manufacturing. In contrast, a machine vision strategy has recently been adopted to distinguish between successful and unsuccessful laser welding items. This opened new perspectives for evaluating the quality of the laser welding product using digital image techniques. However, these methods cannot deliver outstanding performance across multiple laser-welding products and achieve optimal classification accuracy. Deep learning has evolved remarkably in recent years and gained popularity in detecting welding defects. This paper presents the observation of a Hybrid Vision Transformer (HViT) to classify the laser welding images corresponding to their feature patch images based on multi-model feature aggregation. The VGG-16 and MobileNet, which were pre-trained on ImageNet, were utilized as the core models to extract the rich features of the laser welding image. To integrate these features, the squeeze excitation module (SE) was utilized, and a multi-layer perception (MLP) approach with a label smoothing optimizer was used for classification. To determine the effectiveness of the proposed strategy, a comparative analysis was conducted with a multitude of alternative machine learning and deep learning approaches. The results indicate that the proposed strategy surpasses all other methods in terms of classification accuracy and evaluation metrics on the test dataset.

A Comparison of CNN-based Image Feature Extractors for Weld Defects Classification

<span lang="EN-US">Classification of the types of weld defects is one of the stages of evaluating radiographic images, which is an essential step in controlling the quality of welded joints in materials. By automating the weld defects classification based on deep learning and the CNN architecture, it is possible to overcome the limitations of visually or manually evaluating radiographic images. Good accuracy in classification models for weld defects requires the availability of sufficient datasets. In reality, however, the radiographic image dataset accessible to the public is limited and imbalanced between classes. Consequently, simple image cropping and augmentation techniques are implemented during the data preparation stage. To construct a weld defect classification model, we proposed to utilize the transfer learning method by employing a pre-trained CNN architecture as a feature extractor, including DenseNet201, InceptionV3, MobileNetV2, NASNetMobile, ResNet50V2, VGG16, VGG19, and Xception, which are linked to a simple classification model based on multilayer perceptron. The test results indicate that the three best classification models were obtained by using the DenseNet201 feature extractor with a test accuracy value of 100%, followed by ResNet50V2 and InceptionV3 with an accuracy of 99.17%. These outcomes are better compared to state-of-the-art classification models with a maximum of six classes of defects. The research findings may assist radiography experts in evaluating radiographic images more accurately and efficiently.</span>

Microstructures and properties of 7075 aluminum alloy and CP780 steel resistance spot welded joint assisted by magnetic field

Joining steel and aluminum is vital for lightweight automobile but still challenging due to their different physical properties. Herein, resistance spot welding tests were performed on CP780 high-strength steel (thickness 1 mm) and 7075 aluminum alloy (thickness 1.5 mm) dissimilar metals under steady-state magnetic field. The influences of magnetic field (B = 40 mT) on the structure of welded joints, the phase composition/content of intermetallic compounds, and tensile properties of welded joints were analyzed under different welding current conditions (I = 9 kA,10 kA, 11 kA, and 12 kA). At the same welding current, the Lorentz force generated by the additional magnetic field promoted the outward circumferential movement of the molten metal in the weld along the horizontal surface , as well as increased the diameter of the Fe/Al contact interface in the weld nugget along the horizontal direction, conducive to the effective utilization of heat of the resistance spot welding. Except under (11 kA-0 mT) and (11 kA-40 mT), welded joints under other welding parameters displayed a few welding defects, such as incomplete fusion and shrinkage cavity formed at the cross-section of the welded joints. Therefore, the synergism between the magnetic field and appropriate welding current held important roles in the formation of welded joints without obvious welding defects. The intermetallic compounds of all the welds were mainly composed of (Fe, Si)Al2 and (Fe, Si)Al5. Meanwhile, the thickness and content of the intermetallic compounds layer reduced under a magnetic field at the same welding current, significantly improving the tensile properties of the welded joints. The comprehensive properties of welded joints were the best under 11 kA-40 mT, with an average shear force increase from 3.02 kN to 3.49 kN (15.56%) and an average displacement increase from 1.01 mm to 1.22 mm (20.79%). Overall, the proposed dissimilar aluminum/steel resistance spot welded joint assisted by magnetic field looks promising for lightweight automobile use.

Mechanism of pore evolution in electron beam welding joints of Mo-14Re alloy

The mechanism of pore evolution in vacuum electron beam joints of Mo-14Re alloy has been thoroughly investigated. Microscopic characterization methods, including Optical Microscopy (OM) and Scanning Electron Microscopy (SEM), were employed to explore the evolution mechanism of pore defects at different locations. Weld defects were further characterized using Energy Dispersive Spectroscopy (EDS) and Transmission Electron Microscopy (TEM), providing insights into their formation mechanisms. The findings reveal that pores along grain boundaries are the result of MoO and CO bubble aggregation. Non-equilibrium solidification of the melt triggers chemical reactions, transforming MoO and CO bubbles into MoO3 and C, with lower melting points. MoO3 and C persist in the pores, leading to the formation of grain boundary gas hole defects. On the other hand, intragranular pore defects stem from an abundance of pores within the matrix material. Upon entry of gas from these pores into the molten pool, its escape is impeded due to the intense influence of the molten pool. After the molten pool completes the cooling and solidification process, the gas remains within the crystal, resulting in intragranular pore defects.

Enhanced impact toughness of 316L stainless steel welded joint by ultrasonic impact

316 L welded structures have been widely used in the industrial field, but the weld defects seriously affect the reliability. The ultrasonic impact treatment (UIT) is used to improve the quality of welded joints and the effects of the UIT on the impact toughness, fracture morphology, microhardness and microstructure of the 316 L stainless steel welded joint were studied. Moreover, the coupled finite element model of the 316 L stainless steel welding-UIT was established to study the effects of the UIT on the residual stress. The results showed that the UIT could effectively improve the impact toughness of the 316 L stainless steel welded joint, and the impact toughness could be improved with the increasing of the UIT current. According to the fracture morphology, it was found that the fibrous region increased, and the radical region decreased. In addition, the dimple of the material tends to become deeper and larger, and the increase of dimple was accompanied by the increase of energy absorption. The microhardness of the welded joint was improved, and the grain was refined, which indicated that the 316 L stainless steel welded joint generated obvious plastic deformation. The simulation results show the welded joints can introduce the significant residual compressive stress under the action of the UIT. Based on the comprehensive analysis of the UIT on impact toughness, the generated plastic deformation, the reduced grain size, and the introduced residual compressive stress of the 316 L stainless steel welded joint were the main reasons for the improvement of the impact toughness.

Deep convolutional neural network for weld defect classification in radiographic images

The quality of welds is critical to the safety of structures in construction, so early detection of irregularities is crucial. Advances in machine vision inspection technologies, such as deep learning models, have improved the detection of weld defects. This paper presents a new CNN model based on ResNet50 to classify four types of weld defects in radiographic images: crack, pore, non-penetration, and no defect. Stratified cross-validation, data augmentation, and regularization were used to improve generalization and avoid over-fitting. The model was tested on three datasets, RIAWELC, GDXray, and a private dataset of low image quality, obtaining an accuracy of 98.75 %, 90.255 %, and 75.83 %, respectively. The model proposed in this paper achieves high accuracies on different datasets and constitutes a valuable tool to improve the efficiency and effectiveness of quality control processes in the welding industry. Moreover, experimental tests show that the proposed approach performs well on even low-resolution images.

Studying an improved recursive rough method for three-dimensional visual examination of steel plate butt welding surface quality

An improved recursive rough algorithm is proposed to extract feature points in order to address the slow speed and poor accuracy of feature point extraction in 3D visual inspection, as well as the difficulty in meeting the requirements of efficient and quantitative analysis of welded surface quality detection. The preprocessed point cloud is sliced by section. The feature points are obtained according to the improved recursive rough algorithm. The defect evaluation is carried out to obtain the inspection conclusion of the weld appearance defect. Finally, according to the formulated weld defect evaluation process and standards, the typical weld template is selected for weld width, weld misedge and weld straightness test. Weld detection accuracy can reach 3 decimal places, the speed is 4 times the current manual detection speed. The detection results show that the system has the characteristics of high accuracy and fast speed which can replace manual detection. It also has a good application prospect.

Optimization of Welding Parameters for Minimizing Defects and Porosity in Electric Arc Welding of Mild Steel Grade Fe 500 Using of Taguchi Method

Abstract: Electric arc welding is a widely used technique in industrial applications for joining metals, including mild steel grade Fe 500. However, the process is susceptible to defects such as porosity, which can compromise the integrity and quality of welds. In this study, we investigate the optimization of welding parameters to minimize defects and porosity in electric arc welding of mild steel grade Fe 500, utilizing the Taguchi method. The Taguchi method, known for its efficiency in optimization studies, is employed to systematically identify the most influential welding parameters and their optimal levels. Through a series of experiments based on the Taguchi orthogonal array design, we vary parameters such as current amperage, voltage, and electrode size to explore their effects on weld quality. The experimental results reveal significant insights into the relationship between welding parameters and defect formation. By analyzing signal-to-noise ratios and ultimate tensile strength, we quantify the effects of parameter variations on weld quality. The Taguchi analysis enables us to identify the optimal combination of welding parameters that minimizes defects and porosity while maximizing weld strength.

Advanced Optical Coherence Tomography for Real-Time Detection of Defects in Aluminum Alloy Laser Welding

Advanced Optical Coherence Tomography for Real-Time Detection of Defects in Aluminum Alloy Laser Welding

A sequential modelling approach to determine process capability space during laser welding of high-strength Aluminium alloys

Remote laser welding (RLW) technology has become a prominent joining technology in automotive industries, offering high production throughput and cost-effectiveness. Recent advancements in RLW processes such as beam oscillation have led to an increased number of input process parameters, enabling precise control over the heat input to weld metallic materials. A critical necessity in laser welding entails selecting robust process parameters that satisfy all weld quality indicators or key performance indicators (KPIs) during two stages: production stage (often implemented as robotic welding); and repair/rework stage (implemented as cobotic/manual welding to identify process parameters for weld defects) as addressing these factors in both stages is necessary to satisfy near-zero-defect strategy for some e-mobility products.. This research presents a comprehensive methodology that encompasses the following key elements: (i) the development of physics-based simulations to establish the correlation between KPIs and process parameters; (ii) the integration of a sequential modelling approach that strikes a balance between accuracy and computation time to survey the parameter space; and (iii) development of the process capability space for the quick selection of robust process parameters.Three physical phenomena are considered in the development of numerical models, which are (i) heat transfer, (ii) fluid flow and (iii) material diffusion to investigate the effect of process parameters on the weld thermal cycle, solidification parameters and solute intermixing layer during laser welding of dissimilar high-strength aluminium alloys. The governing physical phenomena are decoupled sequentially, and KPIs are estimated based on the governing phenomena. At each step, the process capability space is defined over the parameters space based on the constraints specific to the current physical phenomena. The process capability space is determined by the constraints based on the KPIs. The process capability space provides the initial combination of process parameter space during the early design stage, which satisfies all the KPIs, thus decreasing the number of experiments. The proposed methodology provides a unique capability to (i) simulate the effect of process variation as generated by the manufacturing process, (ii) model quality requirements with multiple and coupled quality requirements, and (iii) optimise process parameters under competing quality requirements.

Synthetic data augmentation for high-resolution X-ray welding defect detection and classification based on a small number of real samples

Deep learning has become the dominant technology in most computer vision tasks. These methods often rely on a large number of labeled sample datasets for training, and in the field of non-destructive testing of welds in industrial manufacturing, weld images with defects are very scarce, and it is still a challenging challenge to construct high-resolution weld defect datasets that meet the requirements. To overcome this limitation, a new data augmentation method for high-resolution X-ray welding defect classification and synthesis based on a small number of real samples is proposed to realize the data augmentation of industrial nondestructive inspection X-ray film defect images. Firstly, to overcome the scarcity of the weld X-ray defect classification dataset, the weld X-ray defect classification dataset (Weld Defect Classification, WDC) is constructed. Secondly, the performance of 16 common deep classification models on WDC datasets is explored. Then, the images of the real local welding defects and the non-defective weld area are fused at random locations, and two data augmentation modes, (Single Image Single Defect, SISD) and (Single Image Multi Defects, SIMD), can generate defect files and annotation files (Visual Object Classes, VOC) at the same time, which can save a lot of time for manual marking. Finally, compared with the traditional data augmentation method, the proposed method can effectively improve the accuracy of defect detection and generalization, the mAP (Mean Average Precision, mAP) @0.5 of YOLOV8X (You Only Look Once, YOLO) and YOLOV5.6.1X is 66.6% and 72.8%, which provides an effective solution for data sample generation in the industrial field.

Diagnosis of Al-CFRTP TA-FSLW defect using acoustic emission signal based on SPWVD and ResNet

As one of the important development directions in modern railway and automobile, lightweight requires reliable connection technology of dissimilar materials. Thermal-assisted friction stir lap welding (TA-FSLW) is an available welding technology for the joining of dissimilar materials, which can solve the problems of physical and chemical properties in the welding process of dissimilar materials. Unfortunately, defects may occur during the welding process. This paper proposes a defect diagnosis method based on smooth pseudo Wigner-Ville distribution (SPWVD) and ResNet18 for Al-CFRTP TA-FSLW. Firstly, the acoustic emission (AE) signal and the temperature of the weld nugget are acquired by the AE sensor and thermography camera in the FSLW process. Then, the time–frequency domain features of AE signal are extracted by SPWVD, and the maximum temperature of the weld nugget is extracted by exponential moving average (EMA) from thermography. Moreover, the dataset is established by SPWVD features, maximum temperature of welding nugget, and several time–frequency features. Finally, the ResNet18-attention network is employed to classify welding defects and stages. The Al-CFRTP TA-FSLW experiments are carried out based on the proposed method. Experimental result indicates that the SPWVD spectrum can effectively extract the feature frequency of deep and shallow hole defects in the welding process. In the section of classification, the accuracy of the dataset with SPWVD features is improved by 2% to 4% compared to other methods, which demonstrates that the SPWVD feature is superior in defect feature extraction.

Non-destructive detection and analysis of weld defects in dissimilar pulsed GMAW and FSW joints of aluminium castings and plates through 3D X-ray computed tomography

AbstractThis work focuses on porosity formation during the welding of dissimilar aluminium alloys (cast and sheet) by pulsed gas metal arc welding (GMAW) with different travel speeds (12–14 mm/s) and by friction stir welding (FSW). The case study concerns the assembling of a battery-pack enclosure prototype. The welded specimens were scanned by 3D X-ray computed tomography. The cast base material (BM) shows a porosity percentage of 1.45%, and it is characterized by pores with a strong hyperbolic relationship between equivalent diameter and sphericity. Considering the GMAW beads, porosity rises with the travel speed (from 1.80 to 5.12%), due to the reduction of the opening window in which pores can escape. Pores with volume higher than 0.10 mm3 rise with the travel speed, representing from 9.75 to 32.98% of the total porosity. These pores are responsible for the weaker hyperbolic connection for sphericity found for the GMAW beads. On the other hand, FSW mixes and homogenizes the pores in the cast BM. The novelty of the paper lays in proving the strong potentialities of FSW for weld porosity reduction. A re-designing of the battery-pack enclosures is necessary to limit arc welding in marginal areas, which are not crucial for sealing but necessary to create a stable platform to be subsequently sealed with FSW.

Preliminary study on terahertz non-destructive testing for defect detection in hot melt joints of polyethylene pipes

With the rapid development of the economy and society, non-metallic pipelines are more and more widely used in oil and gas fields, water supply projects, urban gas systems, and so on. Among the conventional nondestructive testing (NDT) methods, only ultrasonic, radiographic, and penetration testing tools are applicable to non-metallic pipe examination from a theoretical perspective, while their actual efficiency is quite low. In this work, terahertz (THz) non-destructive detection of typical welding defects in polyethylene-pipe hot melt joints is carried out. The characteristic THz wave signal is obtained, and the relationship between the THz wave and the defect type is established. Finally, a THz signal processing model for defect detection in polyethylene-pipe hot melt joints is established. Thus, this research lays a foundation for the application of the THz non-destructive testing technology to polyethylene pipeline engineering.

Multiobjective optimization of laser welding parameters for P92 steel based on MFO/MOEAD-Kriging

P92 steel is used in steam equipment of thermal power plants and nuclear power plants. It needs to withstand high temperature and high pressure during service, so high-quality welds are required. Laser welding can easily obtain good weld quality and reduce the probability of coarse grains and welding defects in the heat-affected zone. Post-weld deformation, residual stress, and weld hardness are important indexes for evaluating weld quality, which largely depends on the combination of laser welding process parameters. The multiobjective optimization methods of moth-flame optimization (MFO) and multiobjective evolutionary algorithm based on decomposition under the Kriging model were compared in this research. It was found that the MFO algorithm has a faster optimization speed and better overall effect. The Optimal Latin hypercube sampling method was used to design the simulation test. The welding seam-forming process was simulated by the SYSWELD software. The Kriging model was used to make the nonlinear relationship between the process parameters and the weld quality indexes significant. A new data set was designed to evaluate the accuracy of the model. On this basis, the efficiency of the combination of process parameters optimized by the two algorithms was compared. The simulation and experimental results were in good agreement with the multiobjective optimization results of the MFO. In the experimental verification, the optimized weld quality was analyzed from the aspects of microstructure and mechanical properties. The results showed that the multiobjective optimization method quickly and effectively reduces the probability of weld quality defects in actual welding, thereby improving weld quality. This multiobjective optimization method provides valuable research ideas for engineering research and development to effectively shorten its development cycle.

The Pre-Heating Effect for Porosity Control during the Laser Welding of Galvanized Steel Sheets

Recently, for the high performance of automobiles, the application ratio of zinc-coated steel sheets with excellent corrosion resistance has been increasing. However, it is difficult to achieve sound welds, as porosities form from zinc during welding. In this study, a laser welding process, with the addition of a pretreatment step to conventional laser welding, was devised to improve weld quality by eliminating pores caused by zinc in the welds. Laser welding was performed on the overlapping joint of hot-dip galvanized steel sheets (SGARC 340, Hyundai Steel., Suncheon-Si, Jeollanam-do, Republic of Korea), with a thickness of 1.2 mm in the upper sheet and 0.9 mm in the lower sheet, and the welding characteristics were investigated by varying the laser power and focal position. Compared to conventional laser welding, the three-step laser stitch welding process significantly reduced the degree of spatter generation and welding defects. Additionally, it increased the tensile–shear load by approximately 37%. Moreover, the reduction in the zinc component of the three-step laser stitch welds was confirmed through SEM-EDS analysis. These findings contribute valuable information for securing high-quality welding joints and effectively addressing zinc-coated steel sheet welding quality issues in the automotive and industrial sectors.

The Influence of Welding Technique With Variation of Current on Surface Defects and Penetration of ST40 Steel

One of the connection of one place starting from several materials or materials that are solid is one of the welding methods, as for the factors that can affect the quality of the weld results which include the process of tools and materials to be used, welding direction, welding electrode movement patterns, in order to use and select electrodes. In this study ST 40 steel plate, testing was carried out with 3 specimens using U and zigzag movement patterns and the direction of welding used back and forth. variations in electric current used are 40 A, 60 A, 80 A. This study has the aim of influencing welding techniques using current variations. The test results that the welding process using welding techniques consisting of movement patterns, welding direction and welding time does not affect the weld defects in the welding results. Welds defects that occur due to variations in the current used

Application of Improved Convolutional Neural Network in Defect Identification of Exhaust Pipe Welds

<p>This article focuses on the identification of welding defects in engine exhaust pipe welds. Firstly, a binocular vision system is built, and the models and parameters of the cameras and lenses involved in the entire system are explained in detail. At the same time, the cameras are calibrated; Then, in response to the problems of large volume, low efficiency, and lack of attention mechanism in the current neural network model, the network model was improved by adding MP structure, CA attention mechanism, and other methods to improve the recognition efficiency of the model. Finally, the reliability of the proposed method was verified through simulation experiments, and the overall recognition efficiency was improved to 97.28%.</p> <p> </p>

Microstructure, Variant Selection, and Mechanical Properties of Laser-Welded Ti-4Al-2V Joints

Laser welding of the near α-phase titanium alloy Ti-4Al-2V, used for complex components in the nuclear industry, has been rarely reported. In this study, butt weld joints made of Ti-4Al-2V alloy plates under different parameters, including the laser power, the welding speed, and the defocus distance, were manufactured and analyzed. The results showed that adjusting the combination of 4.2 kW of laser power, a 20 mm/s welding speed, and a −2 mm defocus distance could achieve a penetration depth exceeding 6 mm. Porosity defects were prone to forming in the middle and bottom parts of the fusion zone, due to rapid cooling. The microstructure of the fusion zone was mainly needle-like α martensite, which precipitated in the form of specific clusters. The interior of a cluster was composed of three types of variants with <11−20>/60° phase interfaces to achieve the lower boundary’s energy. Affected by the microstructure and welding defects, the strength of the weld joint was basically similar under different welding conditions, namely about 720 MPa, slightly higher than that of the base metal, while the rupture elongation at breaking decreased by more than 50%. The micro-Vickers hardness of the weld joints was about 50–60 HV higher than that of the base metal, while the impact toughness was about 40 KJ, almost half that of the base metal. This research lays a solid foundation for the engineering application of laser welding of Ti-4Al-2V alloys.

Study on improvement of weld defect in oscillating laser welding of aluminum alloy T-joints assisted by solder patch

During oscillating laser welding of aluminum alloy T-joints assisted by solder patch, it is found that the pore defect can be reduced and the high quality welded T-joints are obtained. To understand the effect of oscillating laser on the formation process of T-joint welds, a dynamic model for laser welding of aluminum alloy T-joints assisted by solder patch is developed in this paper to analyze the dynamic behaviors of keyhole and molten pool. Based on the dynamic model, the morphology characteristics of keyhole and the evolution processes of molten pool temperature and flow fields during conventional laser welding and oscillating laser welding are calculated and compared. The simulation results are found to be kept consistent with the experimental results. It is demonstrated that the keyhole is characterized with better morphology and higher stability and the degassing condition of molten pool is improved during oscillating laser welding compared with conventional laser welding. The molten pool temperature and flow fields are much more uniform during oscillating laser welding. The obtained results are efficient for suppressing weld defects and achieving high quality welding of aluminum alloy T-joints.