Quality control is a critical component of industrial production, particularly in the aviation industry, where precision and reliability are paramount. This research addresses the challenge of automating the inspection process for X-ray images of engine blade castings using deep learning-based image augmentation techniques. A significant obstacle in this field is the limited availability of training data, especially defective samples, which hinders the development of robust classification models. To overcome this challenge, a prototype image representing the entire dataset was introduced as a baseline for evaluating the fidelity of augmented images. Various augmentation methods were analyzed using Minkowski distance metrics, with Euclidean distance selected as the most effective measure. Principal Component Analysis (PCA) was employed to visualize the relationship between augmented and original images. Experimental results demonstrated that the optimal augmentation settings involved limiting translation and shear transformations while maintaining a Fill Value equal to the mean pixel intensity of the original dataset. These findings contribute to the development of more reliable deep learning models for industrial quality control applications. The proposed approach enhances dataset diversity, improving defect detection accuracy and enabling more effective automated inspection processes in non-destructive testing scenarios.