Application of Deep Learning Algorithms in Studies of Cutting Tool Degradation using Image Processing

Abstract

In the industrial industry, it is crucial to keep an eye on how cutting tools are deteriorating. High-quality products in terms of geometry, residual stress, and surface finish are not produced by tools that are heavily worn. Additionally, inefficient tool replacement might result in higher production costs and lost productivity. Thus, keeping an eye on the tool's health is crucial to preventing these extra expenses and guaranteeing high-quality output. In particular, VGG19, EfficientNetV2, and Vision Transformers are among the categorization models examined in this paper. These models use the tools' images to categorize their condition. The top-performing AI-based image analysis models are compared using transfer learning to determine which are best suited for cutting tool monitoring. They are compared in terms of explainability, performance, and generalizability. With an accuracy of 94%, VGG19 is the model that performs the best, followed by ViT and EfficientNetV2, both of which have accuracys of 87%. A thorough comparison of these findings is done.