Burrs and Sharp Edge Detection of Metal Workpiece Using CNN Image Classification Method for Intelligent Manufacturing Application

This study aims to establish a deep learning method for the real-time and accurate detection of sharp edges and burrs in metal workpieces. This paper used the Single Shot MultiBox Detector (SSD) network which was combined with the VGG16 convolutional neural network (CNN) to form the VGG16-SSD. This method was utilized as the meta-structure for the proposed surface defect detection method. The structure of the SSD was optimized to enhance its accuracy, and the network structure and parameters were refined to simplify the detection process. The proposed approach was tested on an aluminum specimen using the CNN program to detect sharpness and burrs along its edges. The results showed that the method could accurately detect surface defects with a 98.75% accuracy rate. These results provide new insights into defect detection in actual industrial settings and contribute to the advancement of this field. The model underwent a process of tuning to optimize its predictive capability.