Detecting balling defects using multisource transfer learning in wire arc additive manufacturing

Abstract

Wire arc additive manufacturing (WAAM) has gained attention as a feasible process in large-scale metal additive manufacturing due to its high deposition rate, cost efficiency, and material diversity. However, WAAM induces a degree of uncertainty in the process stability and the part quality owing to its non-equilibrium thermal cycles and layer-by-layer stacking mechanism. Anomaly detection is, therefore, necessary for the quality monitoring of the parts. Most relevant studies have applied machine learning to derive data-driven models that detect defects through feature and pattern learning. However, acquiring sufficient data is time- and/or resource-intensive, which introduces a challenge to applying machine learning-based anomaly detection. This study proposes a multi-source transfer learning method that generates anomaly detection models for balling defect detection, thus ensuring quality monitoring in WAAM. The proposed method uses convolutional neural network (CNN) models to extract sufficient image features from multi-source materials, then transfers and fine-tunes the models for anomaly detection in the target material. Stepwise learning is applied to extract image features sequentially from individual source materials, and composite learning is employed to assign the optimal frozen ratio for converging transferred and present features. Experiments were performed using a gas tungsten arc welding-based WAAM process to validate the classification accuracy of the models using low-carbon steel, stainless steel, and Inconel.

Effects of Contaminations on Electric Arc Behavior and Occurrence of Defects in Wire Arc Additive Manufacturing of 316L-Si Stainless Steel

“…Process parameters potentially inducing defects encompass the interplay between the wire feed rate and torch speed, heat input, Contact Tip to Work Distance (CTWD), and gas flow rate [2]. These causative factors may lead to the occurrence of defects, including porosity, voids, fissures, deformations, a lack of fusion, oxidation, and delamination [12][13][14], which demand avoidance, particularly in components subjected to extreme environments, where such defects may precipitate failure mechanisms, such as elevated-temperature fatigue [15]. Given that in the GMAW-based WAAM process the electric current directly influences the material deposition [16], there exists a heightened susceptibility to issues such as spattering, porosity, and excessive heating, relative to WAAM processes employing GTAW and PAW techniques [17].…”

“…Furthermore, Zhang et al [30] proposed a technique for monitoring the quality of weld beads in WAAM utilizing electrical signals, employing a Swin transformer model to construct a classification model capable of accurately distinguishing between regular weld beads and those exhibiting surface oxidation defects. Shin et al [12] detailed a machine learning framework for identifying and categorizing welder porosity defects using welding voltage signals and X-ray imagery. The proposed approach involved training neural network models on X-ray images with and without porosity defects to facilitate detecting and classifying such defects.…”