China Mechanical Engineering ›› 2025, Vol. 36 ›› Issue (12): 2936-2943.DOI: 10.3969/j.issn.1004-132X.2025.12.016

Previous Articles    

Tool Wear Monitoring Based on IWOA-IECA-BiLSTM Model

Zhenke BAO(), Huajun CAO(), Fengze QIN, Zhixiang CHEN, Guibao TAO   

  1. State Key Laboratory of Mechanical Transmission,Chongqing University,Chongqing,400044
  • Received:2025-01-18 Online:2025-12-25 Published:2025-12-31
  • Contact: Huajun CAO

基于IWOA-IECA-BiLSTM模型的刀具磨损监测

包振科(), 曹华军(), 秦逢泽, 陈志祥, 陶桂宝   

  1. 重庆大学机械传动国家重点实验室, 重庆, 400044
  • 通讯作者: 曹华军
  • 作者简介:包振科,男,2000年生,硕士研究生。研究方向为铣削加工刀具磨损状态监测。E-mail:2698566059@qq.com
    曹华军*(通信作者),男,1978年生,教授、博士研究生导师。研究方向为绿色制造战略及基础理论,数控机床绿色化、智能化,产品及制造系统碳排放核算与精益管控,智能基础零部件。E-mail: hjcao@cqu.edu.cn
  • 基金资助:
    国家重点研发计划(2022YFB3206700)

Abstract:

To improve the monitoring accuracy of tool wear during machining, a BiLSTM model based on IWOA and IECA mechanism was proposed. Tool wear data segments from the PHM2010 dataset were intercepted, and multi-domain features were extracted. Tool wear strongly correlated features were then obtained by screening with the Pearson correlation coefficient. The input features were used to train the model. The BiLSTM module in the model effectively captured temporal features within the data. The IECA attention mechanism module enhances the feature representational capability. The IWOA module optimized the model's hyperparameters, further improving the model accuracy. The model performance was finally tested based on three-fold cross-validation and compared with several other models. The results demonstrate that the IWOA-IECA-BiLSTM tool wear monitoring model achieves the best performance on most test sets. On test sets C1C4 and C6, the root mean square error (RMSE) values are as low as 6.5, 12.46, and 9.28, respectively.

Key words: tool wear, improved whale optimization algorithm(IWOA), improved efficient channel attention(IECA) mechanism, bi-directional long short-term memory(BiLSTM) network

摘要:

为了提高加工过程中刀具磨损监测精度,提出一种基于改进的鲸鱼优化算法(IWOA)和改进的高效通道注意力机制(IECA)的双向长短期记忆网络(BiLSTM)模型。通过对PHM2010刀具磨损数据进行片段截取并提取多域特征,再结合皮尔逊系数筛选得到刀具磨损强相关特征。输入特征训练模型,模型中BiLSTM模块能有效捕捉数据中的时序特征;IECA注意力机制模块能提高特征表征能力;IWOA模块能优化模型超参数,进一步提高模型精度。最后基于三折交叉验证测试模型性能,并与其他多个模型进行对比,结果表明,IWOA-IECA-BiLSTM刀具磨损监测模型在多数测试集上具有最佳表现,在C1C4C6三个测试集上均方根误差分别低至6.5、12.46、9.28。

关键词: 刀具磨损, 改进鲸鱼优化算法, 改进高效通道注意力机制, 双向长短期记忆网络

CLC Number: