In this study, the authors propose an integrated sensor-based framework consisting of two core components: a robot control module and a welding quality inspection module. The proposed system relies on vision sensors to acquire real-time visual information from the brazing process. For robot control, image-based sensing using a vision sensor and You Only Look Once-based object detection are performed to enhance positional accuracy and adaptability during autonomous brazing processes. For quality assessment, a convolutional neural network combined with a temporal attention mechanism is utilized to capture both spatial and temporal characteristics of the welding process, enabling the robust classification of weld quality. Experimental results demonstrate that the proposed approach achieves an F1-score of 98% under target manufacturing conditions. These findings highlight the potential of deep-learning-based vision and attention mechanisms for improving process reliability and automation in intelligent manufacturing environments.