In engineering structures, welded joints are the most critical and prominent parts. These welded joints are used in various industries such as shipbuilding, Automobile, aerospace, railroads, construction for holding different parts of the structure together. Non-destructive evaluation (NDE) techniques play a crucial role in finding the quality of weld defects such as porosity, slag inclusions, cracks, and incomplete penetration. Detection and classification of such weld defects are cumbersome and require domain expertise. Artificial Intelligence (AI) is gaining a lot of attention in recent times and can play a significant role in object detection and classification of weld defects, thereby reducing human intervention. Developing a reliable and efficient AI-based automated defect recognition (ADR) system requires a large annotated training dataset, but obtaining such a dataset is challenging in the NDE domain. In this work, we have proposed using an AI algorithm for automating the process of large synthetic Total Focusing Method (TFM) imaging dataset creation using a small set of Finite Element (FE) simulation datasets. This technique permits the generation of datasets that are several orders faster when compared to the FE method.