AI summaryⓘ
The authors developed SimPhysNet, a new method to predict how deeply a laser weld will penetrate, which is important for making strong welds. Their method needs only a small set of labeled images by using a special learning technique that teaches the computer to understand physical aspects of welding from a larger set of unlabeled images. This approach combines physics knowledge with smart image tricks to help the model learn better and then use only a few labeled examples to classify weld penetration accurately. They showed that SimPhysNet performs almost as well as traditional methods that require many more labeled examples. This work could help automate and improve the quality control of laser welding.
Laser weldingFull-penetrationSelf-supervised learningPhysics-informed neural network (PINN)Contrastive learningFew-shot learningPrototypical networksImage augmentationMolten poolKeyhole
Authors
Sen Li, Xiaoying Liu, Xiaojian Xu, Chendong Shao, Yaqi Wang, Ling Lan, Xinhua Tang, Haichao Cui
Abstract
The laser welding full-penetration is of critical importance, as it constitutes one of the fundamental factors in achieving defect-free welded joints. Accurate prediction of the penetration state is therefore essential for ensuring weld quality. To this end, this paper introduces SimPhysNet, a novel algorithm that achieves high classification accuracy in laser welding penetration prediction using only a limited number of labelled images. This approach effectively overcomes the limitations of supervised learning classification algorithms, which are hindered in industrial applications by their dependence on extensive, high-quality labelled data. The core of SimPhysNet is a unique self-supervised learning paradigm that embeds physical priors into a contrastive learning framework. By incorporating a physics-informed neural network (PINN), the model is guided to extract physically meaningful features of the molten pool and keyhole from a large set of unlabelled data, while three image augmentation tasks further enhance its generalization capabilities. Subsequently, a few-shot learning strategy, based on prototypical networks, enables robust classification by constructing class representations from a minimal set of labelled images. Experimental results demonstrate that SimPhysNet achieves a classification accuracy of 96.06% using only 200 labelled images (approximately 5% of the total labelled dataset), which is comparable to the performance of conventional supervised learning algorithms that utilize the entire labelled dataset. This work presents a new, efficient, and highly accurate method, providing the way for the intelligent automation of laser welding.