Integrated Fuzzy Logic and Image Processing for Automated Quality Grading in Metal Casting

Improving quality control in high-volume manufacturing remains a critical challenge, especially where manual inspections are error-prone and inconsistent. This study presents an automated quality assessment framework combining image processing and fuzzy logic techniques to enhance defect detection and classification. A dataset comprising 1,300 grayscale images of disk-shaped cast products (512×512 pixels) was analyzed. Gaussian filtering and Canny edge detection were applied to extract defect features such as size, color deviation, and shape distortion. These features were processed through a fuzzy inference system to assign quality scores, categorizing products into low (≤40), medium (41–70), and high quality (>70). Results revealed that 84.8% of products were low quality, while only 3.6% were rated high quality. Additionally, the Fuzzy TOPSIS method was implemented to rank products based on similarity to an ideal solution, offering a decision-support perspective beyond classification. This hybrid approach improves consistency, objectivity, and scalability in quality evaluation workflows.