改进VMD-LSTM法在刀具磨损状态识别中的应用

姜超; 李国富

doi:10.13433/j.cnki.1003-8728.20200627

改进VMD-LSTM法在刀具磨损状态识别中的应用

doi: 10.13433/j.cnki.1003-8728.20200627

姜超^1,,
李国富^2, ,

1.
宁波大学机械工程与力学学院, 浙江宁波 315211
2.
宁波大学先进储能技术与研究院, 浙江宁波 315211

基金项目:

国家自然科学基金项目 51705263

详细信息

作者简介:
姜超(1997-), 硕士研究生, 研究方向为信号处理与故障诊断, 947030121@qq.com

通讯作者:
李国富, 教授, 硕士生导师, liguofu@nbu.edu.cn

中图分类号: TH164
计量
- 文章访问数: 55
- HTML全文浏览量: 8
- PDF下载量: 15
- 被引次数: 0
出版历程
- 收稿日期: 2021-05-21
- 刊出日期: 2022-02-25

Application of Modified VMD and LSTM in Tool Wear State Recognition Model

JIANG Chao^1
,,
LI Guofu^{2
, ,}

1.
Faculty of Mechanical Engineering & Mechanics, Ningbo University, Ningbo 315211, Zhejiang, China
2.
Institute of Advanced Energy Storage Technology and Equipment, Ningbo University, Ningbo 315211, Zhejiang, China

摘要

摘要: 针对车刀在实际加工时工况复杂导致磨损状态识别精度不高的问题，提出了一种基于最大包络峰度法的变分模态分解(Variational mode decomposition，VMD)结合长短时记忆网络(Long short-term memory，LSTM)的组合分类算法。采用最大包络峰度法确定VMD最佳分解模态数，计算信噪比对高频信号进行降噪重构，然后对原始信号以及分解后的信号进行特征提取和清洗，针对数据样本不均衡的问题，引入SMOTE算法合成少数类样本，结合特征变化以及刀具加工过程中的磨损划分数据集，使用LSTM模型实现多工况下车刀磨损状态的分类。最后通过实验验证所提出的模型和方法的有效性，实验结果表明，此模型与其他分类模型相比具有更高的分类精度以及更好的泛化性。
- 刀具状态监测 /
- 电流信号 /
- 变分模态分解 /
- 特征提取 /
- 长短时记忆网络
Abstract: Aiming at the problem of low accuracy of wear state recognition due to complex working conditions of turning tools in actual machining, a combined classification algorithm of variational mode decomposition (VMD) based on the maximum envelope kurtosis combined with long and short-term memory (LSTM) networks is proposed in this paper. First, the maximum envelope kurtosis method is used to determine the optimal decomposition mode number of VMD, the signal-to-noise ratio is calculated to reduce the noise and reconstruct the high-frequency signal, and then the original signal and the decomposed signal are feature extraction and cleaning. Then, aiming at the unbalanced data sample for the problem, the SMOTE algorithm is introduced to synthesize a minority of samples, combined with feature changes and the wear division data set during tool processing, and the LSTM model is used to classify the wear status of the turning tool under multiple working conditions. Finally, the effectiveness of the proposed model and method is verified through experiments. The experimental results show that this model has higher classification accuracy and better generalization than other classification models.
- tool state monitoring /
- current signal /
- variational mode decomposition (VMD) /
- feature extraction /
- long and short-term memory (LSTM)

HTML全文

图 1 试验流程图

下载: 全尺寸图片幻灯片

图 2 车刀信号与频谱图

下载: 全尺寸图片幻灯片

图 3 最大包络峰度变化趋势图

下载: 全尺寸图片幻灯片

图 4 VMD分解波形图

下载: 全尺寸图片幻灯片

图 5 去噪后信号

下载: 全尺寸图片幻灯片

图 6 特征变化趋势图

下载: 全尺寸图片幻灯片

图 7 模型精度训练和损失曲线

下载: 全尺寸图片幻灯片

表 1 试验切削条件表

组号	转速/ (r·min^-1)	切削深度/mm	进给速度/ (mm·r^-1)
1	480	1.0	0.176
2	760	1.0	0.176
2	760	1.5	0.192
3	1 080	1.0	0.176

下载: 导出CSV

表 2 实验样本数量表

组号	1	2	3	4
样本数量	273	244	190	214

下载: 导出CSV

表 3 LSTM模型结构

LSTM模型结构	各层神经元数量	各层激励函数	各层输入维度
输入层	特征维度	无	(None, 1, 特征维度)
LSTM层	30	Sigmoid	(None, 30)
全连接层	8	Relu	(None, 8)
输出层	3	Softmax	(None, 3)

下载: 导出CSV

表 4 模型测试结果

分类模型	测试集平均分类准确率/%	测试集最优分类准确率/%	训练时间/s
LSTM	93.78	94.02	83.35
PSO-SVM	90.43	91.67	89.49
BP神经网络	88.59	89.21	39.86

下载: 导出CSV

参考文献(16)

[1]	ZHANG Y, ZHU K P, DUAN X Y, et al. Tool wear estimation and life prognostics in milling: model extension and generalization[J]. Mechanical Systems and Signal Processing, 2021, 155: 107617 doi: 10.1016/j.ymssp.2021.107617
[2]	GOWTHAMAN P S, JEYAKUMAR S, SARAVANAN B A. Machinability and tool wear mechanism of Duplex stainless steel-a review[J]. Materials Today: Proceedings, 2020, 26: 1423-1429 doi: 10.1016/j.matpr.2020.02.295
[3]	袁军, 刘丽冰, 张艳蕊, 等. 刀具磨损状况的检测方法研究综述[J]. 现代制造工程, 2021(3): 152-160 https://www.cnki.com.cn/Article/CJFDTOTAL-XXGY202103027.htm YUAN J, LIU L B, ZHANG Y R, et al. Survey of research on detection methods for tool wear condition[J]. Modern Manufacturing Engineering, 2021(3): 152-160 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-XXGY202103027.htm
[4]	陈刚, 焦黎, 颜培, 等. 基于多传感器数据融合的刀具磨损状态监测研究[J]. 新技术新工艺, 2017(11): 23-28 https://www.cnki.com.cn/Article/CJFDTOTAL-XJXG201711007.htm CHEN G, JIAO L, YAN P, et al. Research on tool wear condition monitoring based on multi-sensor data fusion[J]. New Technology & New Process, 2017(11): 23-28 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-XJXG201711007.htm
[5]	LI W J, LIU T S. Time varying and condition adaptive hidden Markov model for tool wear state estimation and remaining useful life prediction in micro-milling[J]. Mechanical Systems and Signal Processing, 2019, 131: 689-702 doi: 10.1016/j.ymssp.2019.06.021
[6]	戴稳, 张超勇, 孟磊磊, 等. 基于深度学习与特征后处理的支持向量机铣刀磨损预测模型[J]. 计算机集成制造系统, 2020, 26(9): 2331-2343 https://www.cnki.com.cn/Article/CJFDTOTAL-JSJJ202009003.htm DAI W, ZHAO C Y, MENG L L, et al. Support vector machine milling wear prediction model based on deep learning and feature re-processing[J]. Computer Integrated Manufacturing Systems, 2020, 26(9): 2331-2343 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-JSJJ202009003.htm
[7]	赵明利, 袁一鸣, 宋士杰, 等. 基于特征选择和模糊支持向量机的刀具磨损状态识别[J]. 制造技术与机床, 2020(11): 115-120 https://www.cnki.com.cn/Article/CJFDTOTAL-ZJYC202011025.htm ZHAO M L, YUAN Y M, SONG S J, et al. Tools wear state recognition based on feature selection and fuzzy support vector machine[J]. Manufacturing Technology & Machine Tool, 2020(11): 115-120 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZJYC202011025.htm
[8]	DRAGOMIRETSKIY K, ZOSSO D. Variational mode decomposition[J]. IEEE Transactions on Signal Processing, 2014, 62(3): 531-544 doi: 10.1109/TSP.2013.2288675
[9]	郑圆, 胡建中, 贾民平, 等. 一种基于参数优化变分模态分解的滚动轴承故障特征提取方法[J]. 振动与冲击, 2020, 39(21): 195-202 https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCJ202021027.htm ZHENG Y, HU J Z, JIA M P, et al. A method for rolling bearing fault feature extraction based on parametric optimization VMD[J]. Journal of Vibration and Shock, 2020, 39(21): 195-202 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-ZDCJ202021027.htm
[10]	李帅永, 夏传强, 程振华, 等. 基于VMD和互谱分析的供水管道泄漏定位方法[J]. 仪器仪表学报, 2019, 40(7): 195-205 https://www.cnki.com.cn/Article/CJFDTOTAL-YQXB201907023.htm LI S Y, XIA C Q, CHENG Z H, et al. Leak location method in water-supply pipeline based on combination of VMD and cross-spectrum analysis[J]. Chinese Journal of Scientific Instrument, 2019, 40(7): 195-205 (in Chinese) https://www.cnki.com.cn/Article/CJFDTOTAL-YQXB201907023.htm
[11]	马洪斌, 佟庆彬, 张亚男. 优化参数的变分模态分解在滚动轴承故障诊断中的应用[J]. 中国机械工程, 2018, 29(4): 390-397 doi: 10.3969/j.issn.1004-132X.2018.04.003 MA H B, TONG Q B, ZHANG Y N. Applications of optimization parameters VMD to fault diagnosis of rolling bearings[J]. China Mechanical Engineering, 2018, 29(4): 390-397 (in Chinese) doi: 10.3969/j.issn.1004-132X.2018.04.003
[12]	周怡娜, 董宏丽, 张勇, 等. 基于VMD去噪和散布熵的管道信号特征提取方法[J/OL]. 吉林大学学报(工学版), 1-13. [2021-05-16]. https://doi.org/10.13229/j.cnki.jdxbgxb20200889 ZHOU Y N, DONG H L, ZHANG Y, et al. Feature extraction method of pipeline signals based on VMD de-noising and dispersion entropy[J/OL]. Journal of Jilin University (Engineering and Technology Edition), 1-13. [2021-05-16]. https://doi.org/10.13229/j.cnki.jdxbgxb20200889 (in Chinese)
[13]	PARWAL V, ROUT B K. Machine learning based approach for process supervision to predict tool wear during machining[J]. Procedia CIRP, 2021, 98: 133-138 doi: 10.1016/j.procir.2021.01.018
[14]	王毫. 基于LSTM循环神经网络的BTA钻头磨损监测技术研究[D]. 西安: 西安理工大学, 2019 WANG H. Research on BTA drill bit wear monitoring technology based on LSTM recurrent neural network[D]. Xi'an: Xi'an University Of Technology, 2019 (in Chinese)
[15]	CAI W L, ZHANG W J, HU X F, et al. A hybrid information model based on long short-term memory network for tool condition monitoring[J]. Journal of Intelligent Manufacturing, 2020, 31(6): 1497-1510 doi: 10.1007/s10845-019-01526-4
[16]	WU X Q, LI J, JIN Y Q, et al. Modeling and analysis of tool wear prediction based on SVD and BiLSTM[J]. The International Journal of Advanced Manufacturing Technology, 2020, 106(9-10): 4391-4399 doi: 10.1007/s00170-019-04916-3