Scholarlink

<p>Accurate real-time prediction of tool wear is crucial for dynamically optimizing cutting parameters. However, many existing studies overlook the dynamic adaptability between tool wear and cutting parameters, which may compromise machining quality and increase the energy consumption of machine tools. To address the aforementioned issues, this paper proposes a framework for dynamic cutting parameter optimization using online tool wear prediction as input. Firstly, a high-dimensional nonlinear feature dimensionality reduction algorithm based on arc-cosine kernel singular value decomposition (AC-KSVD) was proposed, and an AC-KSVD-PSO-BiLSTM tool wear prediction model was established by combining the particle swarm algorithm (PSO) and the bi-directional long-short-term memory network (BiLSTM). Secondly, a mapping model linking tool wear to surface roughness and cutting energy consumption was established using a BP neural network. Under specified constraints, an improved cuckoo search algorithm (TS-CS) incorporating a tabu memory mechanism was employed to optimize cutting parameters that vary with wear conditions. Experimental results demonstrate that the proposed AC-KSVD-PSO-BiLSTM tool wear prediction model achieves an accuracy rate of 95%. The optimized cutting parameters derived from this model reduce machine tool energy consumption by 53.66% while maintaining machining quality.</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

Dynamic cutting parameter optimization based on deep learning online tool wear monitoring model

  • 期刊论文

  • 发表日期：

    2026年5月23日

查看全文
  • Guochao Qiao,
  • Yongxiang Zhu,
  • Dong Zhen

摘要

Accurate real-time prediction of tool wear is crucial for dynamically optimizing cutting parameters. However, many existing studies overlook the dynamic adaptability between tool wear and cutting parameters, which may compromise machining quality and increase the energy consumption of machine tools. To address the aforementioned issues, this paper proposes a framework for dynamic cutting parameter optimization using online tool wear prediction as input. Firstly, a high-dimensional nonlinear feature dimensionality reduction algorithm based on arc-cosine kernel singular value decomposition (AC-KSVD) was proposed, and an AC-KSVD-PSO-BiLSTM tool wear prediction model was established by combining the particle swarm algorithm (PSO) and the bi-directional long-short-term memory network (BiLSTM). Secondly, a mapping model linking tool wear to surface roughness and cutting energy consumption was established using a BP neural network. Under specified constraints, an improved cuckoo search algorithm (TS-CS) incorporating a tabu memory mechanism was employed to optimize cutting parameters that vary with wear conditions. Experimental results demonstrate that the proposed AC-KSVD-PSO-BiLSTM tool wear prediction model achieves an accuracy rate of 95%. The optimized cutting parameters derived from this model reduce machine tool energy consumption by 53.66% while maintaining machining quality.