Scholarlink

<p>Tool wear has an important impact on machining product quality and productivity. Since tool wear data are usually characterized by nonlinearity, complexity and small sample size, the traditional single prediction model cannot fully meet the higher prediction requirements. In order to overcome this problem, a&#xa0;multi-sensor tool wear prediction model based on Logistic-improved Hippopotamus Optimization-Back Propagation network (Logistic-improved HO-BP) is given in this paper, which optimizes the HO algorithm to compute the optimal weights and thresholds of the BP network by using the good traversal, cross-correlation and autocorrelation characteristics of the Logistic graph. The vibration and cutting force signals derived from the PHM (Predictive and Health Management) Association’s 2010 CNC Machine Tool Health Prediction Competition’s dataset are preprocessed using time-domain segmentation, Hampel filtering, and wavelet denoising. Subsequently, time-domain, frequency-domain and time-frequency-domain features are extracted from the preprocessed data and analyzed by Pearson correlation and XGBoost feature importance analysis as model inputs. In this paper, the public data released by PHM 2010 is used for comparative implementation, and the experimental results have successfully proved the superiority of XGBoost in feature selection as well as usability and high accuracy of Logistic-improved HO-BP model for tool wear prediction.</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

Tool wear prediction system based on multi-sensor fusion and Logistic-improved HO-BP optimization

  • 期刊论文

  • 发表日期：

    2025年8月6日

查看全文
  • Chunhua Zhao,
  • Mingxin Zhao,
  • Jinling Tan,
  • Qian Li,
  • Zhangwen Lin,
  • Jian Tang,
  • Qiyang Wang,
  • Xinyi Hu,
  • Shengyong Wang,
  • Minghang Yang

摘要

Tool wear has an important impact on machining product quality and productivity. Since tool wear data are usually characterized by nonlinearity, complexity and small sample size, the traditional single prediction model cannot fully meet the higher prediction requirements. In order to overcome this problem, a multi-sensor tool wear prediction model based on Logistic-improved Hippopotamus Optimization-Back Propagation network (Logistic-improved HO-BP) is given in this paper, which optimizes the HO algorithm to compute the optimal weights and thresholds of the BP network by using the good traversal, cross-correlation and autocorrelation characteristics of the Logistic graph. The vibration and cutting force signals derived from the PHM (Predictive and Health Management) Association’s 2010 CNC Machine Tool Health Prediction Competition’s dataset are preprocessed using time-domain segmentation, Hampel filtering, and wavelet denoising. Subsequently, time-domain, frequency-domain and time-frequency-domain features are extracted from the preprocessed data and analyzed by Pearson correlation and XGBoost feature importance analysis as model inputs. In this paper, the public data released by PHM 2010 is used for comparative implementation, and the experimental results have successfully proved the superiority of XGBoost in feature selection as well as usability and high accuracy of Logistic-improved HO-BP model for tool wear prediction.