The integrity of welded joints is a crucial influence in the field of structural engineering, as it has significant implications for the dependability and safety of diverse sectors, including aerospace, automotive, and construction. In order to ensure the optimal performance of welded joints, it is important to evaluate their microstructural features. This paper explains a novel methodology for investigating the integrity of welded joints by employing microstructural characterization with the use of a deep encoder. The deep encoder is a neural network design that has exhibited exceptional skills in applications related to image processing and feature extraction. In the course of our study, we modified this particular framework to examine microstructural pictures of welded joints, facilitating the acquisition of significant information pertaining to their structural soundness. The methodology employed in this study entails the acquisition of high-resolution images of welded connections, often obtained by the utilization of scanning electron microscopy or comparable imaging techniques. The findings of our investigation exhibit promise, as they offer a more precise and effective approach for evaluating the integrity of welded joints in contrast to conventional manual techniques.