Image Preprocessing Method in Radiographic Inspection for Automatic Detection of Ship Welding Defects

Welding defects must be inspected to verify that the welds meet the requirements of ship welded joints, and in welding defect inspection, among nondestructive inspections, radiographic inspection is widely applied during the production process. To perform nondestructive inspection, the completed weldment must be transported to the nondestructive inspection station, which is expensive; consequently, automation of welding defect detection is required. Recently, at several processing sites of companies, continuous attempts are being made to combine deep learning to detect defects more accurately. Preprocessing for welding defects in radiographic inspection images should be prioritized to automatically detect welding defects using deep learning during radiographic nondestructive inspection. In this study, by analyzing the pixel values, we developed an image preprocessing method that can integrate the defect features. After maximizing the contrast between the defect and background in radiographic through CLAHE (contrast-limited adaptive histogram equalization), denoising (noise removal), thresholding (threshold processing), and concatenation were sequentially performed. The improvement in detection performance due to preprocessing was verified by comparing the results of the application of the algorithm on raw images, typical preprocessed images, and preprocessed images. The mAP for the training data and test data was 84.9% and 51.2% for the preprocessed image learning model, whereas 82.0% and 43.5% for the typical preprocessed image learning model and 78.0%, 40.8% for the raw image learning model. Object detection algorithm technology is developed every year, and the mAP is improving by approximately 3% to 10%. This study achieved a comparable performance improvement by only preprocessing with data.

IDENTIFICATION OF INCOMPLETE PENETRATION WELDING DEFECTS ON DIGITAL RADIOGRAPHIC FILM FIGURES WITH THE GEOMETRIC INVARIANT MOMENT

The examination of defects in radiographic films necessitates specialized knowledge, as indicated by an expert radiographer (AR) degree, yet the subjectivity of AR in identifying defects is problematic. To overcome this subjectivity, an automatic welding defect identification is needed. This is executed by using Matlab to create artificial neural networks, which is beneficial for users with the graphical user interface (GUI) feature. One of the breakthroughs in the figure extraction into seven feature vector values is the geometric invariant moment theory. This prevents translation, rotation, and scaling from changing the figure's characteristics. Therefore, a welding defect identification system with a geometric invariant moment was created in the digital radiographic film figure to overcome the reading error by AR. The identification system obtained an accuracy rating of 89.9%.

Dissimilar Rotary Friction Welding of Inconel 718 to F22 Using Inconel 625 Interlayer

Dissimilar metal joining has always been a challenging task because of the metallurgical incompatibility and difference in melting points of alloys being joined. Diffusion and mixing of alloying elements from dissimilar base metals at the weld often cause unwanted metallurgical changes resulting in unsuccessful welds or underperformance of the weldment. Solid-state dissimilar friction welds of Inconel 718 and F22 were prepared in this study with an Inconel 625 interlayer to address the carbon enrichment of Inconel 718 during the welding. Defect-free rotary friction welds were produced in this study. Microstructural and mechanical properties investigation of the weldments and base metals was carried out, and results were analysed. Intermixing zone was observed at the weld interface due to the softening of the metal at the interface and rotatory motion during the welding. The high temperatures and the plastic deformation of the intermixing zone and thermo-mechanically affected zone (TMAZ) resulted in the grain refinement of the weld region. The highest hardness was observed at the Inconel 718/F22 weld interface due to the plastic strain and the carbon diffusion. The tensile specimens failed in the F22 base metal for the weld prepared with and without the Inconel 625 interlayer. Inconel 718/F22 welds exhibited lower toughness values compared to the Inconel 718/F22 welds prepared with Inconel 625 interlayer.

Efficient Welding Defect Detectionand Classification of Phased Array C-Scan Images Using Modified Fast Fuzzy C Means Andradial Bias Function Neural Network

In the recent past, phased array technology is one of the most important methodologies used for inspection of welding. The welding defect identification is a difficult task due to noise content and uneven illumination and contrast on phased array 2D image. Artificial Neural Network (ANN) is a recent Machine Learning (ML) technology that has been achieved a lot of attention over the recent years. The saliency feature extraction for representing image has become complex due to quality of 2D image. The proper image restoration and enhancement techniques should be applied in order to improve the quality of 2D phased array image. The 2D-Adaptive Anisotropic Diffusion Filter (2D AADF) is applied to eliminate noises such as impulse noise and speckle noise. The Adaptive Mean Adjustment-Contrast Limited Adaptive Histogram Equalization (AMA-CLAHE) is the enhancement technique that is applied to improve contrast and brightness of the phased array 2D image. The welding defect region can be exactly segmented using saliency mapping to contour boundaries of defects in welding. In this paper, a novel methodology for welding defect detection is applied based on Modified Fast Fuzzy C Means (MFFCM) clustering technique by integrating Probability Mass Function (PMF) threshold technique for higher range of efficient and accurate segmentation. The Gray Level Co-Occurrence Matrix (GLCM) and 2D Band-let Transform (2D BT) are applied to extract features on segmented image. TheRadial Bias Function Neural Network (RBFNN) classifier is one of the ANN classifier for classifying welding defects. Most of image classification techniques utilize RBFNN as they will provide great range of accuracy and precision while compared to existing techniques. The localized generation error model is implemented in RBFNN in order to minimize Mean Square Error (MSE). The efficiency and accuracy of the proposed methodology has been evaluated with the help of experimental results in terms of graphical representation and numerical analysis.

Physical Mechanism of Laser Excited Acoustic Wave and its Application in Recognition of Incomplete-penetration Welding Defect

Incomplete penetration is a type of welding defect that severely impacts the quality of weldments. In order to identify penetration levels in pulsed laser and plasma transferred arc (laser-PTA) hybrid welding, this paper uses structure-borne acoustic sensors to detect acoustic signals. Their characteristics are then analyzed with respect to the time and frequency domains. Acoustic signals characteristic of incomplete-penetration defects were extracted using a Butterworth band-pass filter. Physical mechanisms of laser excited acoustic wave were then studied by analyzing the correlation between incomplete-penetration defects and their characteristic acoustic signals. The results showed that acoustic signals correlating to incomplete-penetration defects have characteristic frequencies ranging from 0 to 10 kHz, which are generated by interaction between the pulsed laser beam and molten pool. An incomplete-penetration defect constitutes an acoustic cavity, which is an acoustic transmission structure. The structure of phonation sources and the acoustic cavity are affected by levels of penetration, giving rise to acoustic signals with different characteristics. In general, the study of physical mechanisms of laser excited acoustic wave lays a foundation for on-line identification of incomplete-penetration defects in laser-PTA hybrid welding.