AI-Enabled Quality Control: Detecting Defects in Manufacturing Processes with Convolutional Neural Networks

This paper aims to establish an improvement upon previous approaches to defect detection, namely, the CNN model and how it is more effective than edge detection and histogram-based methods for manufacturing defects. Overall, the model that used a set of 10,000 images as the training basis presented an accuracy rate of 98. 5%, precision of 98. 7%, recall of 98. 65%, accuracy of 97. 85%, precision of 97. 22%, recall of 3 %, and an F1-score of 98. Only 5%, which is much better than using traditional methods such as multiple linear regressions. The outlined flexibility in relation to the given form of defects and manufacturing environment proves the model’s applicability and efficiency for the purpose of improving quality assurance, which makes it possible asserting its effectiveness. As much as this new advanced AI technology has some limitations especially when it has to do with data demands and computation, the potential that comes with this technology is phenomenal, especially when it comes to defect detection and increased operational efficiency. Based on these observations, the technical suggestions for future improvements are as follows: transfer learning approach, and edge computing to support the model in keeping up optimized learning continually in different manufacturing scenarios.