Automatic Detection of Defective Metal Pressed Components Using Artificial Intelligence

The inspection process for metal pressed components before shipment involves time-consuming manual checks for various defects. In this paper, we introduce a technique to automate the identification of flawed regions on the surface of a component. This is accomplished through the utilization of the Artificial Intelligence object detection framework. The proposed approach offers efficient and reliable defect detection by leveraging AI's high accuracy and real-time processing capabilities. The methodology involves image preprocessing, model training, performance evaluation, and real-time system deployment. Experimental conducted to evaluation on a dataset of 1268 images comprising both defective and non-defective components demonstrate an impressive accuracy of 99.272℅. The processing time required for defect detection is reduced to 4 seconds.