A hybrid deep learning framework for intelligent casting defect detection and quality assurance in manufacturing

Casting is a vital manufacturing process; however, defects in cast products can result in compromised quality, higher production costs, and reduced customer satisfaction. To overcome these challenges, this work developed a novel deep learning-based hybrid framework for intelligent defect detection in casting images. In this, a stacked Convolutional Neural Network (CNN) is employed to extract low-level spatial features. To capture global contextual information and complex feature dependencies, a Transformer model is integrated into the architecture. Furthermore, Batch Normalization is incorporated to stabilize training, accelerate convergence, and reduce the risk of overfitting. The model is trained and validated on a Kaggle dataset comprising 7,348 images, categorized as defective or non-defective. The proposed hybrid model achieved a testing accuracy of 99.02%, demonstrating superior performance compared to existing state-of-the-art methods. Precision, Recall, and F1-Score results confirm the framework’s robustness in reliable casting defect analysis.