Scholarlink

<p>This review introduces a three‑layer architecture (perception, edge processing, and cloud‑based decision) to re‑evaluate digital tool wear monitoring in milling. The analysis covers the main wear mechanisms (flank, crater, edge wear), the role of cemented carbide substrates and PVD/CVD coatings, and the evolution from traditional offline measurement to integrated digital systems. Direct measurement methods (optical microscopy, SEM, profilometry) are compared with indirect sensor‑based approaches that use cutting forces, vibration, acoustic emission, temperature, and computer vision. Emphasis is placed on digital monitoring architectures, examining data acquisition (perception), local processing (edge computing), and cloud‑based decision making. The review assesses the growing application of artificial intelligence (AI) and machine learning algorithms – including convolutional neural networks (CNNs), long short‑term memory networks (LSTMs), and digital twins – for real‑time wear prediction and classification. Based on an analysis of the state-of-the-art, the findings indicate that while accuracy has improved, real‑time robustness, generalization, and uncertainty quantification remain key research frontiers. A central challenge identified here is the velocity at which wear must be identified and quantified, driven by the accelerating pace of digitalization, Industry 4.0 and 5.0, and the growing demand for rapid industrial decision‑making. This imperative is redefining the core objective of wear monitoring toward near‑real‑time responsiveness. Accordingly, the review outlines forward‑looking pathways that prioritize physics‑informed hybrid modeling, open benchmark datasets and validation protocols, edge‑cloud architectures, and explainable AI, thereby paving the way from laboratory validation toward reliable, high‑speed industrial deployment aligned with the requirements of modern smart manufacturing ecosystems.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A review of tool wear monitoring in milling: perception, edge processing and cloud decision

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月23日

查看全文
  • Marcel Henrique Militão Dib,
  • João Paulo Davim

摘要

This review introduces a three‑layer architecture (perception, edge processing, and cloud‑based decision) to re‑evaluate digital tool wear monitoring in milling. The analysis covers the main wear mechanisms (flank, crater, edge wear), the role of cemented carbide substrates and PVD/CVD coatings, and the evolution from traditional offline measurement to integrated digital systems. Direct measurement methods (optical microscopy, SEM, profilometry) are compared with indirect sensor‑based approaches that use cutting forces, vibration, acoustic emission, temperature, and computer vision. Emphasis is placed on digital monitoring architectures, examining data acquisition (perception), local processing (edge computing), and cloud‑based decision making. The review assesses the growing application of artificial intelligence (AI) and machine learning algorithms – including convolutional neural networks (CNNs), long short‑term memory networks (LSTMs), and digital twins – for real‑time wear prediction and classification. Based on an analysis of the state-of-the-art, the findings indicate that while accuracy has improved, real‑time robustness, generalization, and uncertainty quantification remain key research frontiers. A central challenge identified here is the velocity at which wear must be identified and quantified, driven by the accelerating pace of digitalization, Industry 4.0 and 5.0, and the growing demand for rapid industrial decision‑making. This imperative is redefining the core objective of wear monitoring toward near‑real‑time responsiveness. Accordingly, the review outlines forward‑looking pathways that prioritize physics‑informed hybrid modeling, open benchmark datasets and validation protocols, edge‑cloud architectures, and explainable AI, thereby paving the way from laboratory validation toward reliable, high‑speed industrial deployment aligned with the requirements of modern smart manufacturing ecosystems.