Sagar H. Nikam

Recent advances in deep convolutional neural networks (D-CNNs) have improved automated welding defect inspection. This study presents an explainable comparative framework for multi-class classification of defects in Aluminium 5083 TIG weld joints using High Dynamic Range (HDR) image data, integrating a transfer-learning model, stratified five-fold cross-validation, computational-time analysis, and Grad-CAM-based visual interpretation. Five transfer-learning-based D-CNN architectures such as VGG16, VGG19, Inception V3, MobileNet, and DenseNet were trained, validated, and tested under a common evaluation protocol to assess their suitability for welding defect classification. The dataset was organised into classes such as good weld, contamination, lack of fusion, lack of penetration, and misalignment. Model performance was compared using multiple evaluation metrics. Stratified five-fold cross-validation was also performed to assess model stability. Alongside the cross-validation, training/inference times were also recorded to evaluate computational feasibility. Grad-CAM was used as an explainable artificial intelligence (XAI) technique in order to provide visual interpretation of weld regions. Among evaluated models, DenseNet achieved the best overall performance, with a classification accuracy of 98%, and showed the least confusion across defect classes. The Grad-CAM visualisations showed that the model focused on defect-relevant weld regions, demonstrating that transfer-learning D-CNNs with XAI can support TIG welding defect classification and effective visual quality assessment.

Spatters in the powder-based metal additive manufacturing processes influence deposition quality, part surface quality, and internal defects. We developed a novel video analysis method to monitor and count the melted spatter particles of biomedical-grade Co-Cr-Mo-4Ti powder particles in depositing layers using a micro-plasma transferred arc additive manufacturing (M-PTAAM) process. We captured the spatters using a weld-monitoring camera and building datasets of videos and monitored different combinations of M-PTAAM process parameters. We captured videos of the melted spatter particles and counted the melted spatter particles in real time using a Kalman filter. The weld-monitoring camera captured the melted spatter particles and the fumes generated by the evaporated spatter particles. The video processing algorithm was developed in this study to accurately capture melted spatter particles. In images without fumes, nearly all melted spatter particles were successfully detected. Even in images with the presence of fumes, the algorithm maintained a detection accuracy of 90%. The real-time melted spatter count particle exhibited a powder feed rate changing from 30 to 35 g/min and then to 50 g/min. The melted spatter particle count was lowest at a powder feed rate of 30 g/min and increased with the increasing powder feed rate.

Reducing the defects in the additively manufactured components using Laser-Directed Energy Deposition (L-DED) process is important for ensuring structural integrity, surface quality, and functional performance. The first required step for reducing defects in the L-DED manufactured components is the identification and understanding of the type of defects using the object detection approach. This paper aims to use a YOLO-based object detection models to classify and detect defects in the horizontal wall, vertical wall, and cuboid structures manufactured using various combinations of L-DED process parameters. The objectives involved are training, testing and validating of YOLOv7, YOLOv8, YOLOv9, and YOLOv9-GELAN models on the independent dataset of defects such as flash formation, void and rough texture, identifying the best YOLO model capable of detecting small and big size multiple defects within a single image and comparing the defects captured by YOLO model with previously used conventional CNN model such as VGG16. The results revealed that YOLOv9-GELAN exhibited good performance indicators compared to other YOLO models. The increasing trend for mAP0.5:0.95 signifies YOLOv9-GELAN as a good choice for defect detection of multiple defects in a single image. It also gave mAP of 95.7%, precision of 94%, recall of 96%, and F1-score of 90%, indicating accuracy in defect localisation and classification with minimal false positives and negatives. These high values for YOLOv9-GELAN indicate its capability to accurately highlight the defects using the bounding box compared to the previously proposed VGG16 model. In addition, YOLOv9-GELAN capability of processing 62 images per second showed its potential for higher frames processing compared to other YOLO models. This research will progress the development of AI-based in-situ defect monitoring for the L-DED process.

Defects formed during the spreading of powder, known as powder bed defects, are a major issue in additive manufacturing processes. Deep learning (DL)-based image classification models can be utilised to detect defects caused by the powder spreading process. The aim of this research was to review and compare the performance of the EfficientNet_v2 deep learning image classification model against the commonly used VGG-16 model on a selective laser sintering powder bed defects (SLS PBDs) dataset. It was observed that the EfficientNet_v2 model achieved higher performance than the commonly used VGG-16 model, with a defect prediction accuracy of 97.54% and model sensitivity of 96.3%.

This paper presents a deep learning approach to identify and classify various defects in the laser-directed energy manufactured components. It mainly focuses on the Convolutional Neural Network (CNN) architectures, such as VGG16, AlexNet, GoogLeNet and ResNet to perform the automated classification of defects. The main objectives of this research are to manufacture components using the laser-directed energy deposition process, prepare a dataset of horizontal wall structure, vertical wall structure and cuboid structure with three defective classes such as voids, flash formation, and rough textures, and one non-defective class, use this dataset with a deep learning algorithm to classify the defect and use the efficient algorithm to detect defects. The next objective is to compare the performance parameters of VGG16, AlexNet, GoogLeNet and ResNet used for classifying defects. It has been observed that the best results were obtained when the VGG16 architecture was applied to an augmented dataset. With augmentation, the VGG16 architecture gave a test accuracy of 94.7% and a precision of 80.0%. The recall value is 89.3% and an F1-Score is 89.5%. The VGG16 architecture with augmentation is highly reliable for automating the defect detection process and classifying defects in the laser additive manufactured components.

The properties of Extra Low Interstitials (ELI) Ti6Al4V components fabricated via the laser-based powder bed fusion (L-PBF) process are prone to variation, particularly throughout a powder reuse regime. Interstitial pick-up of interstitial elements within the build chamber during processing can occur, most notably, oxygen, nitrogen, and hydrogen, which can impair the mechanical properties of the built component. This study analyses ELI Ti6Al4V components manufactured by the L-PBF process when subjected to a nine-stage powder reuse sequence. Mechanical properties are reported via hardness measurement and tensile testing. Results showed that from 0.099 wt.% to 0.126 wt.% oxygen content, the mean hardness and tensile strength increased from 367.8 HV to 381.9 HV and from 947.6 Mpa to 1030.7 Mpa, respectively, whereas the ductility (area reduction) reduced from around 10% to 3%. Statistical analysis based on the empirical model from Tabor was performed to determine the strength–hardness relationship. Results revealed a significant direct relationship between tensile strength and Vickers hardness with a proportionality constant of 2.61 (R-square of 0.996 and p-value of 6.57 × 10⁻⁶).

The Laser-based Powder Bed Fusion (L-PBF) process uses a laser beam to selectively melt powder particles deposited in a layer-wise fashion to manufacture components derived from Computer-Aided Design (CAD) information. During laser processing, material is ejected from the melt pool and is known as spatter. Spatter particles can have undesirable geometries for the L-PBF process, thereby compromising the quality of the powder for further reuse. An integral step in any powder replenishing and reuse procedure is the sieving process. The sieving process captures spatter particles within the exposed powder that have a diameter larger than a defined mesh size. This manuscript reports on Ti6Al4V (Grade 23) alloy powder that had been subjected to seven reuse iterations, focusing on the characterisation of powder particles that had been captured (i.e., removed) by the sieving processes. Characterisation included chemical composition focusing upon interstitial elements O, N and H (wt.%), particle morphology and particle size analysis. On review of the compositional analysis, the oxygen contents were 0.43 wt.% and 0.40 wt.% within the 63 µm and 50 µm sieve-captured powder, respectively. Additionally, it was found that a minimum of 79% and 63% of spatter particles were present within the captured powder removed by the 63 µm and 50 µm sieves, respectively.

Processes that use moving point heat sources to temporarily create localized melt pools (metal additive manufacture and fusion welding) have a flow phenomenon due to the surface tension gradient. Surface tension of the liquid metal reduces with temperature and this, coupled with the high temperature gradients associated with point heat sources, creates Marangoni convection in the melt. The Marangoni convection tends to reduce the temperature and change the melt pool geometry (increases width but reduces depth). Computational Fluid Dynamics (CFD) models can simulate the phenomenon of Marangoni convection but are computationally intensive. A simpler thermal model involving heat conduction and latent heat, but with the liquid’s thermal conductivity artificially increased by a constant factor, exhibits similar thermal effects to the Marangoni convection. The heat conduction models are computationally less intensive than CFD, but the trial-and-error exercise needed to obtain an appropriate multiplying factor is time consuming. With an aim to improve the process of factor selection, the present study investigates the correlation between the surface tension gradient and correction factors. For a Ti-6Al-4V under typical additive manufacturing parameters, the corresponding correction factor to be applied to liquid thermal conductivity was 1.76.

Titanium alloy powder used for laser-based powder bed fusion (L-PBF) process is costly. One of the solutions is the inclusion of a powder recycling strategy, allowing unused or exposed powder particles to be recuperated post manufacture, replenished and used for future builds. However, during a L-PBF process, powder particles are exposed to high levels of concentrated energy from the laser. Particularly those in close proximity to the melt pool, leading to the formation of spatter and agglomerated particles. These particles can settle onto the powder bed, which can then influence the particle size distribution and layer uniformity. This study analysed extra-low interstitial (ELI) Ti6Al4V (Grade 23) powder when subjected to nine recycle iterations, tracking powder property variation across the successive recycling stages. Characterisation included chemical composition focusing upon O, N, and H content, particle size distribution, morphology and tapped and bulk densities. On review of the compositional analysis, the oxygen content exceeded the 0.13% limit for the ELI grade after 8 recycles, resulting in the degradation from Grade 23 level.