Weld Defect Detection Research Articles

Welding defect detection based on local image enhancement

Welding defect detection in a radiographic image is an important topic in the field of industrial non-destructive testing. To improve the accuracy of welding defect segmentation, a local image enhancement approach is proposed. In this algorithm, the requirement of contrast enhancement is considered when extracting the weld seam and segmenting the weld defect. The whole defect detection is conducted by three procedures: image enhancement, welding seam extraction, and defect segmentation. Firstly, a method for determining the Localised Pixel Inhomogeneity Factor (LPIF) is proposed. Then, based on the results of LPIF, the Otsu method is applied to segment the welding seam and defects are, identified by region growing algorithm. The authors compared LPIF with histogram equalisation, adaptive histogram equalisation, and contrast-limited adaptive histogram equalisation algorithms and assessed its performance by using indicators such as image contrast, image definition, and edge intensity. Moreover, the authors compared the segmentation results of the enhanced defect images with the original image to further study the method's effect on weld defect segmentation. More than 70 images containing various types of defects are tested. The experimental results demonstrate that the quality of enhanced defect images is improved significantly, and has a high relative segmentation accuracy of more than 92%.

Data-Driven Detection of Laser Welding Defects Based on Real-Time Spectrometer Signals

The spectrometer is applied in this paper to get the in-process information of welding defects during the high-power disk laser welding process. The high-dimensional signal captured by the spectrometer is fed into a data-driven framework based on stacked auto-encoder (SAE) to automatically extract salient features for performing real-time welding defects detection. The particle swarm optimization (PSO) algorithm is employed to optimize the proposed data-driven framework by acquiring the global optimal parameters to strengthen its capability in extracting the representative features from the original high-dimensional signal. The extracted features are classified by a softmax classifier to get the real-time identifications of the welding defects. The proposed framework is compared with the conventional shallow artificial intelligent methods, such as back-propagation (BP) neural network and support vector machine (SVM), and reveals better performance. The actual welding experiments under different welding parameters are implemented to validate the detection accuracy of our proposed data-driven framework. This paper provides an effective framework for the detection of the high-power disk laser welding status in real-time.

A spectroscopic method based on support vector machine and artificial neural network for fiber laser welding defects detection and classification

Diverse welding processes have been utilized in manufacturing industry for years. But up to date, welding quality still cannot be guaranteed, due to the lack of an efficient and on-line welding defects monitoring method, and this leads to increased manufacturing costs. In this paper, a method based on feature extraction and machine learning algorithm for on-line quality monitoring and defects classification was presented. Plasma radiation was captured by an optical fiber probe, and delivered by an optical fiber to the spectrometer. The captured spectral signal was processed by selecting sensitive emission lines and extracting features of spectral data's evolution, which realized spectral data compression with low computational cost. After selecting the proper training data set, the designed ANN and SVM allows automatic detection and classification of welding defects. The validity of proposed method was successfully approved by test data set in welding experiments. Welding experiments on galvanized steel sheets showed the corresponding relationship between the output of classifiers and welding defects. Finally, the two classifiers were compared. Experiments indicated the performance of ANN is slightly better than that of SVM, while both of them have its own advantages.

Detection model of invisible weld defects by magneto-optical imaging at rotating magnetic field directions

Magneto-optical (MO) imaging non-destructive testing (NDT) system excited by rotating magnetic field is proposed for feature extraction and detection classification of invisible arbitrary-angle weld defects. The magnetic field direction of the rotating magnetic field changes periodically with time, and the magnitude and direction of the leakage magnetic field on the weldment also changes periodically. A finite element analysis (FEA) model of weldment is established to study the distribution of rotating alternating current field at different transient time. The correctness of the rotating alternating current field theory is verified by FEA. Based on the Faraday rotation effect, the relationship between the imaging characteristics of weld defect MO image and the leakage magnetic field intensity is analyzed. The gray value of MO image can match the corresponding leakage magnetic field intensity. MO imaging NDT experiments are performed on invisible weld defects, including surface crack, subsurface crack, and non-penetration. The gray-level co-occurrence matrix (GLCM) method is used to extract texture features of the MO image of weld defects, and the texture features of the images can reflect the leakage magnetic field characteristics of the defects. These texture features of MO images are used as input data for the defect classification model established using support vector machine (SVM). This model is evaluated by weld defect detection experiments and shows that it can effectively and accurately classify invisible arbitrary-angle weld defects.

EMD- PNN based welding defects detection using laser-induced plasma electrical signals

The plasma electrical signal has gained extensive attention for characterizing the behavior of the laser-induced plasma due to the advantages of easy acquisition and feedback control. In this paper, the electrical signals were measured by a passive probe based on the principle of plasma sheath effect. To explore the mutation characteristics of plasma electrical signals during defect generation in laser deep penetration welding, wavelet packet transform (WPT) and empirical mode decomposition (EMD) were used to compress data and extract features, respectively. Based on the analysis of the time-frequency spectrum of a typical plasma electrical signal, the approximate coefficients of 0˜390 Hz frequency range were reconstructed. The residual term which characterizes the change trend of electrical signal was obtained by the further adaptive decomposition. For better identifying weld defects, another two statistical features, mean value and standard deviation, were extracted by carrying out statistical analysis in the time domain. The feature database is built with above features and used as inputs of the predictive model based on the probabilistic neural network (PNN). The result showed the average prediction accuracy was as high as 90.16% when recognizing five statuses of weld seam, including sound weld and four kinds of weld defects.

Multidirectional magneto-optical imaging system for weld defects inspection

Magneto-optical (MO) imaging can be used to detect weld defects under alternating magnetic field excitation. However, in an alternating magnetic field, an MO sensor is more sensitive to defects perpendicular to the excitation direction than defects with other directions. This paper introduces a rotating magnetic field generated by an orthogonal inducer and proposes a multidirectional MO imaging nondestructive testing (NDT) system using rotating magnetic field excitation for the detection of natural weld defects. The relationship between imaging characteristics of MO images and magnetic field strength is analyzed based on the Faraday rotation effect. A finite element analysis (FEA) model of cruciform weld cracks is established to study the distribution of rotating magnetic fields at different transient time. MO imaging NDT experiments were performed on weld defects including surface crack, subsurface crack, and non-penetration. The finite element model is proved by MO imaging experiments of cruciform weld cracks. Defect information of the MO images is extracted through fusion image technology. The MO imaging effects of weld defects under alternating magnetic field excitation and rotating magnetic field excitation are compared. The maximum difference of image gray value is analyzed to determine the effect difference between the two magnetic fields. Research results show that MO imaging under rotating magnetic field excitation overcomes the limitation of directional detection of MO imaging under alternating magnetic field excitation and effectively realizes the detection of multidirectional invisible weld defects.

Characterisation of friction stir welds by logistic regression using fractal and wavelet features

ABSTRACTThis paper presents a method for automatic detection of defective welds based on feature extraction using fractal analysis and wavelet transform with Logistic Regression classifier. The proposed methodology consists of pre-processing, Weld seam extraction, and feature extraction followed by classification process. The image obtained from the welded region is converted to grayscale, followed by binarisation under pre-processing. The process of Weld seam extracts the region of interest (ROI), from which features are obtained that describe the texture and irregularities present in the welded image. The simulation study analyses Friction Stir Welding images of two different aluminium grades (AA1100 with AA6061-T6) and that of brass and copper. In the experiments, a Logistic Regression model to perform classification using the features is employed. This is the first work where Logistic Regression has been applied for weld defect detection in Friction Stir Welding. Experimental results show the effectiveness of the proposed classification process which uses less number of powerful features. The classifier has achieved more than 97% accuracy, indicating potential use of the proposed method for the detection of defects in the friction stir welded images.

Weld image deep learning-based on-line defects detection using convolutional neural networks for Al alloy in robotic arc welding

Accurate on-line weld defects detection is still challenging for robotic welding manufacturing due to the complexity of weld defects. This paper studied deep learning–based on-line defects detection for aluminum alloy in robotic arc welding using Convolutional Neural Networks (CNN) and weld images. Firstly, an image acquisition system was developed to simultaneously collect weld images, which can provide more information of the real-time weld images from different angles including top front, top back and back seam. Then, a new CNN classification model with 11 layers based on weld image was designed to identify weld penetration defects. In order to improve the robustness and generalization ability of the CNN model, weld images from different welding current and feeding speed were captured for the CNN model. Based on the actual industry challenges such as the instability of welding arc, the complexity of the welding environment and the random changing of plate gap condition, two kinds of data augmentation including noise adding and image rotation were used to boost the CNN dataset while parameters optimization was carried out. Finally, non-zero pixel method was proposed to quantitatively evaluate and visualize the deep learning features. Furthermore, their physical meaning were clearly explained. Instead of decreasing the interference from arc light as in traditional way, the CNN model has taken full use of those arc lights by combining them in a various way to form the complementary features. Test results shows that the CNN model has better performance than our previous work with the mean classification accuracy of 99.38%. This paper can provide some guidance for on-line detection of manufacturing quality in metal additive manufacturing (AM) and laser welding.

Applicability of eddy current technique in in-bore NDT tool for ITER hydraulic pipe welds

Eddy current testing (ECT) is a potential candidate for non-destructive testing of pipe welds in fusion facilities because of its simple configuration and applicability in radiation environments. This paper provides an applicability assessment of the ECT method to detect weld defects in the ITER blanket module hydraulic pipes. We found that ECT is not only effective for detecting surface defects, but may also be able to detect defects from the opposite side of test objects by optimizing the measurement conditions of the probe (frequency and coil size), especially in cases of cracks or weld imperfections along the weld line.

FATIGUE LIFE ANALYSIS OF RAIL-WELDS USING LINEAR ELASTIC FRACTURE MECHANICS

The initiation and growth of fatigue cracking is mainly due to high stress concentration, heterogeneity and poor quality of weld. The detection and rectification of such weld defects are major concerns of rail network managers to reduce potential risk of rail breaks and derailments. To estimate the fatigue life of welded joints and to analyze the progress of fatigue cracks, a fracture mechanics-based analysis and fatigue models were developed using Finite Element Analysis. The initial flaw is obtained from a sample weld using ultrasonic flaw detecting machine test. Linear Elastic Fracture Mechanics (LEFM) approach based on the Paris law was applied to determine critical crack size and the number of cycles to failure using FRANC3D software. The inspection interval of rail welds before fracture (failure) was suggested based on reliability and life cycle analysis that correspond with minimum overall cost and frequency interval. It is recommended that fracture-based models in combination with reliability analyses can be a sustainable infrastructure decision-making algorithm.

Detection and Cause Analysis of Structural Defects in Lead Clamp Weld of Transformer

In this paper, the causes analysis of the cracking accident at the lead clamp weld-fusion line of transformer in the extra-high voltage substation was conducted.On-site non-destructive detection of weld defects was carried out to determine the size and location of the defects. An anatomical analysis of the clamp sample revealed that there was serious incomplete penetration and incomplete fusion in the internal structure of the weld, which seriously affected the structural strength of the weld. The causes of welding defects were further analyzed, and suggestions for improving product quality and strengthening inspection supervision were put forward.

Random forest-based real-time defect detection of Al alloy in robotic arc welding using optical spectrum

Aluminum alloy of arc welding is the main technology for the key components manufacturing in aerospace, nuclear power, ship and so on. Real-time weld defects detection is still challenging due to the complexity and diversity of weld defects. Arc optical Spectroscopy emission is the key information generated during arc welding process. However, how to select the effective spectrum feature from high dimension of arc spectrum is crucial for improving the accuracy of defects recognition. This paper proposed an on-line defects detection method for aluminum alloy in robotic arc welding based on random forest and arc spectrum. Firstly, preprocessing of arc spectrum was carried out before 50 features were extracted. Then, a quantitative index of feature importance is proposed based on mean decrease accuracy and mean decrease Gini to reduce the feature redundancy. Six spectral features were selected and analyzed in terms of the construction pattern. Furthermore, the defect identification model was established based on random forest and the optimal feature subset. Comparing with RBF and BP models, it can achieve better performance in identifying three typical defects, including incomplete penetration, burn-through and porosity. This paper can provide some guidance for data mining of optical information and intelligent manufacturing.

Welding defects detection based on deep learning with multiple optical sensors during disk laser welding of thick plates

A multi-sensor system, including an auxiliary illumination (AI) visual sensor system, an UVV band visual sensor system, a spectrometer, and two photodiodes, is established to capture signals of the welding status during high-power disk laser welding. The features of visible light and reflected laser light signal were extracted by decomposing the originally captured signals into different frequency bands by wavelet packet decomposition method (WPD). The captured signal of the spectrometer mainly covers the optical wavelength from 400 nm to 900 nm, which was divided into 25 sub-bands to extract the spectrum features of the spectrometer signal by statistical methods. The features of the plume are acquired by the UVV band visual sensor system, and the features of keyhole are extracted captured from the images captured by AI visual sensor system through the digital image processing method. Based on these quantified real-time features of the welding process, a deep learning algorithm based on convolutional neural network (CNN) was developed to detect three different welding defects during high-power disk laser welding. The established deep learning model is compared with the backpropagation neural network (BP) model, and it shows higher accuracy and robustness in detecting welding defects with the multi-sensor system. Its effectiveness was also validated by four welding experiments with different welding parameters.

Fundamental study on on-line image sensing technology of arc welding – development of in-process quality control technology of arc welding

ABSTRACT To achieve both welding productivity and quality, in-process welding quality control technology is required. The authors are developing method to detect welding defects directly during welding by using image sensing technology for production lines. To detect the welding defects directly, observing the welding phenomena just near arc where the base metal is melted and defects should occur is important. Conventionally, high-speed camera with laser external light source was used for observing the welding phenomena just near arc, however the cost is high and the equipment is complex for production lines. In this paper, the observation results and the required time resolution of welding phenomena just near the arc were verified by comparison with two observation method, using only an industrial CMOS camera without using an external light source and using high-speed camera with laser light source. By comparison with observation result, the welding phenomena just near the arc could be observed by using only CMOS camera from backside when the molten pool flowed under the arc. And the required time resolution for observation of the phenomena just near the arc were as follows: fluctuation of the arc: over 100 Hz, fluctuation of the molten pool: over 60 Hz, droplet transfer: over 600 Hz.

Inspection of butt welds for complex surface parts using ultrasonic phased array

Detection of weld defects for complex surface parts has always been a difficult point in ultrasonic testing because the geometry complexity makes it difficult to arrange transducers and determine the propagation paths of acoustic beams. In this paper, the linear friction weld of the engine blade is taken as an example of the butt weld in complex surface parts, and the application of the ultrasonic array testing method is carried out. Firstly, the propagation properties of acoustic waves in the inspection area are analysed based on both the Snell's law and the acoustic pressure reciprocating transmittance (APRT). According to the inspection requirements, this study establishes a full-coverage inspection solution using multi-array transducers. Secondly, the whole inspection area is divided and the wedge parameters in each subarea are iteratively designed. Thirdly, based on the finite element method (FEM), a response simulation model of the ultrasonic array is established to testify the feasibility and validity of the inspection scheme. Lastly, experiments are conducted on the blade specimen welded by linear friction welding (LFW). The inspection results of different weld positions clearly identify the prefabricated crack defects, showing that the proposed method can fulfill the rapid and accurate inspection for the butt weld of complex surface parts.

Challenges in the detection of weld-defects in friction-stir-welding (FSW)

ABSTRACTFriction-Stir-Welding (FSW) involves many critical process variables that affect the overall quality of the weld produced; consequently, it is difficult to accurately predict the most favourable process parameters to acquire high-quality defect-free welds. Quality of an FSW joint is influenced by several crucial factors including welding process parameters, and many others. Since the new invention of FSW process, investigators have researched the influence of process parameters and pin tool profiles on microstructures and mechanical properties with a view to optimise FSW process, thereby keeping the degradation of the weld strength to a minimum. In this study weld defects in Friction-Stir (FS) welded thick (8.13 mm) aerospace grade aluminium alloy (AA 2219-T87) are correlated with weld process parameters from an assortment of a large matrix of weld schedules. In this investigation, a novel FSW process parameter window was established for the effective joining of aerospace grade aluminium which could successfully distinguish several types of weld schedules with characteristic and intrinsic weld-defects. Using experimental results obtained from various schedules, an empirical force index (Effective Force Index), which is a function of the critical weld process parameters is proposed to identify regions of weld schedules.

Segmentation of Weld Defects Using Multiphase Level Set by the Piecewise-Smooth Mumford-Shah Model

This paper deals with the problem of the X-ray image segmentation used to detect weld defects for a non-destructive testing task. In this work we have implemented a multiphase method using the Mumford and Shah piecewise smooth model to extract the size and the texture information of defect and its region. Using this model which is more robust to the noise and less sensitive to the position of the initial curve, we have obtained a maximum of information for several regions in the X-ray image (the geometry of weld beads and defects). Our results show that the model developed can do at the same time the image segmentation and restoration.

Automatic Welding Defect Detection of X-Ray Images by Using Cascade AdaBoost With Penalty Term

Welding defect constitute a great danger to the safe usage of petroleum pipelines. In addition, manual welding defect detection has numerous deficiencies, such as subjectivity and inaccurate estimation of geometric parameters. Thus, we proposed an automatic welding defect detection system for X-ray images. In the system, five typical defects (cracks, lack of penetration, lack of fusion, round defects, and stripy defects) and non-defects were chosen for recognition. There are three stages in the system: defect extraction, defect detection, and defect recognition. In the first stage, background subtraction with an adaptive thresholding method was adopted to identify the potential defects. In the second stage, to extract real defects from the massive number of potential defects, the adaptive cascade boosting (AdaBoost) algorithm was employed for binary classification. Grayscale features and geometric properties of the defects were extracted for the classification. In the third stage, the AdaBoost algorithm was extended for multi-classification. In the process of distinguishing defects from non-defects, a high detection rate is necessary. To ensure the high true positive rate (TPR) and the low false positive rate (FPR), we proposed the cascade AdaBoost algorithm with penalty term. The accuracy of the defect detection was 85.5%, and the TPR was 91.66%. Moreover, three comparison tests of support vector machine (SVM), k-nearest neighbor (KNN), and random forest (RF) were employed to validate the superiority of AdaBoost. The experimental results indicate that the proposed detection system can be effective for defect detection.

Magneto-optical Imaging Detection and Strong Classification of Weld Defects in Rotating Magnetic Field

Magneto-optical Imaging Detection and Strong Classification of Weld Defects in Rotating Magnetic Field

A Deep Artificial Immune System to Detect Weld Defects in DWDI Radiographic Images of Petroleum Pipes

In recent years, research focused on (semi)automatic radiographic inspection methods has gained more attention. The present work proposes a method for detecting defects in radiographic images of welded joints of oil pipes. Real condition images obtained by the double wall double image (DWDI) technique usually present a lower quality when compared with images traditionally considered in many studies reported in the literature. First, the proposed approach detects discontinuities in DWDI radiographic images of welded joints, and then, based on a hybrid paradigm encompassing artificial immune systems (AIS) and deep learning (DL), it classifies each discontinuity as `defect' and `non-defect'. The proposed method performs two phases in the AIS module: early classification (based on negative selection) and evolving classification (based on clonal selection). In both phases, the pattern recognition task is performed using a set of features extracted from each discontinuity through a detector genetically encoded into immune cells. As an attempt to improve the classification performance, DL models (AlexNet and autoencoders) are incorporated aiming to increase the number of extracted features. Experiments performed on a set of 727 discontinuities show that the proposed approach achieves an Fscore of 70.7%, outperforming each of its modules running by themselves: AlexNet with Fscore = 64.86% and AIS with Fscore = 66%. Considering the challenges imposed by real conditions on image acquisition - and the low rates of false negatives -, results demonstrate that the proposed approach can be used to assist in inspection works when dealing with DWDI images.