Scholarlink

<p>Nowadays, Tool Condition Monitoring (TCM) has become an essential tool, particularly in conventional machining processes, for ensuring product quality and process stability. In light of this, this study was conducted to develop an adaptive TCM system that combines Convolutional Neural Networks (CNN) and a Case-Based Reasoning (CBR) algorithm. To achieve this, a dataset comprising wear images of carbide inserts was created by conducting dry turning experiments on AISI 1045 steel. Subsequently, the number of images was augmented by applying image transformation techniques, which aimed to enhance the generalization ability of the hybrid model. The proposed approach achieved a classification accuracy of 98% for the training set and 96% for the validation set, outperforming similar methods in the literature. The system not only classified tool wear into four distinct levels light, moderate, heavy, and extreme but also detected unseen tool conditions, such as Built-Up Edge (BUE), demonstrating the adaptability offered by the integration of CNN with CBR.</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

Development of an adaptive tool condition monitoring system: integration of case-based reasoning with CNN

  • 期刊论文

  • 发表日期：

    2025年1月21日

查看全文
  • Oussama Dahmoune,
  • Ikhlas Meddour,
  • Mohamed Elbah,
  • Mohamed AtmaneYallese,
  • Salim Belhadi

摘要

Nowadays, Tool Condition Monitoring (TCM) has become an essential tool, particularly in conventional machining processes, for ensuring product quality and process stability. In light of this, this study was conducted to develop an adaptive TCM system that combines Convolutional Neural Networks (CNN) and a Case-Based Reasoning (CBR) algorithm. To achieve this, a dataset comprising wear images of carbide inserts was created by conducting dry turning experiments on AISI 1045 steel. Subsequently, the number of images was augmented by applying image transformation techniques, which aimed to enhance the generalization ability of the hybrid model. The proposed approach achieved a classification accuracy of 98% for the training set and 96% for the validation set, outperforming similar methods in the literature. The system not only classified tool wear into four distinct levels light, moderate, heavy, and extreme but also detected unseen tool conditions, such as Built-Up Edge (BUE), demonstrating the adaptability offered by the integration of CNN with CBR.