基于迁移学习的供药装置故障诊断方法

doi:10.12382/bgxb.2022.0767

摘要/Abstract

摘要：

针对多工况下的模块发射药供药装置故障诊断问题,提出一种基于迁移学习和奇异值分解的故障诊断方法。通过奇异值分解对模块药的位移速度数据进行降维和降噪预处理,并提取特征;采用基于TrAdaBoost算法框架的迁移学习方法,综合有限的试验数据和大量的仿真数据,提取有效故障信息,构建多个基故障分类器,并最终集成一个高质量故障分类器。研究结果表明,该方法对多样工况下的故障数据有很好的适应性,在试验数据量较少的情况下,相对于传统机器学习方法可以获得更好的故障识别准确率。

关键词: 供药装置, 迁移学习, 奇异值分解, 故障诊断

Abstract:

To solve the problem of fault diagnosis of the modular charge feeding mechanism under multiple working conditions, a fault diagnosis method based on transfer learning and singular value decomposition (SVD) was proposed. SVD was used for dimensionality reduction and noise reduction as means of preprocessing of the modular charge velocity data and for feature extraction. The transfer learning method based on the TrAdaBoost algorithm framework was adopted to synthesize limited test data and a large amount of simulation data to extract effective fault information. In the information, multiple base fault classifiers were built and integrated into a high-quality fault classifier. The experimental results showed that the proposed method has good adaptability to the fault data under multiple working conditions, which can obtain better diagnosis accuracy compared to the traditional machine learning strategy in the case of limited test data.

Key words: modular charge feeding mechanism, transfer learning, singular value decomposition, fault diagnosis

中图分类号:

TJ303.3

黄文宽, 钱林方, 尹强, 刘太素. 基于迁移学习的供药装置故障诊断方法[J]. 兵工学报, 2023, 44(10): 2964-2974.

HUANG Wenkuan, QIAN Linfang, YIN Qiang, LIU Taisu. Fault Diagnosis Method of Modular Charge Feeding Mechanism Based on Transfer Learning[J]. Acta Armamentarii, 2023, 44(10): 2964-2974.

图/表 20

图1 模块药供药装置示意图

Fig.1 Schematic of modular charge feeding mechanism

图2 模块药供药装置仿真模型

Fig.2 Simulation model of modular charge feeding mechanism

图3 供药装置机械故障曲线

Fig.3 Mechanical failure curve of modular charge feeding mechanism

表1 选药块数与故障情况的组合

Table 1 Combinations of modular charge numbers and faults conditions

组合序号	缩略组合名	分布占比/%
1	1块-失效	0.06
2	1块-变形	0.12
3	1块-健康	27.8
4	4块-失效	0
5	4块-变形	2.1
6	4块-健康	29.2
7	6块-失效	0.07
8	6块-变形	0.03
9	6块-健康	38.1

图4 干扰噪声的功率谱密度和概率密度函数

Fig.4 Power spectral density and probability density function of noise

图5 典型试验样本和仿真样本速度曲线

Fig.5 Velocity curves of typical simulation and test samples

图6 典型样本奇异值向量的对数曲线

Fig.6 Logarithmic curves of singular value vectors of typical samples

图7 供药装置故障诊断方法流程图

Fig.7 Flowchart of fault diagnosis method for modular charge feeding mechanism

图8 供药装置台架实物图

Fig.8 Bench of modular charge feeding mechanism

图9 储药仓变形局部细节

Fig.9 Details of cartridge deformation

图10 阻尼块磨损局部细节

Fig.10 Details of damping block wear

表2 试验集各类组合样本的占比

Table 2 Distribution of various combinations in the bench test set

组合	样本占比/%	组合	样本占比/%	组合	样本占比/%
1	5	4	5	7	5
2	5	5	5	8	5
3	20	6	20	9	30

表3 仿真集各类组合样本的占比(策略1)

Table 3 Distribution of various combinations in the simulation set (Strategy 1)

组合	样本占比/%	组合	样本占比/%	组合	样本占比/%
1	5	4	5	7	5
2	5	5	5	8	5
3	20	6	20	9	30

表4 仿真集各类组合样本的占比(策略2)

Table 4 Distribution of various combinations in the simulation set (Strategy 2)

组合	样本占比/%	组合	样本占比/%	组合	样本占比/%
1	10	4	10	7	10
2	10	5	10	8	10
3	10	6	10	9	20

表5 仿真样本按策略1分布时的迁移阈值

Table 5 Transfer threshold of Strategy 1

测试错误率/ %	试验∶仿真= 1∶1	试验∶仿真= 1∶2	试验∶仿真= 1∶5
≤20	0.738	0.742	0.713
≤10	0.906	0.852	0.822
≤5			0.927

表6 仿真样本按策略2分布时的最低迁移阈值

Table 6 Transfer threshold of Strategy 2

测试错误率/ %	试验∶仿真= 1∶1	试验∶仿真= 1∶2	试验∶仿真= 1∶5
≤20	0.713	0.681	0.693
≤10	0.881	0.815	0.779
≤5			0.895

图11 3个分类器的混淆矩阵

Fig.11 Confusion matrices of the three classifiers

表7 分药数目分类的准确率和F1值

Table 7 Accuracy and F1 value statistics for modular charge numbers

分药块数	分类器	准确率/%	F1值
	SVM分类器	92.9	0.919
1块药	基分类器	97.0	0.970
	集成分类器	100	1.000
	SVM分类器	86.9	0.864
4块药	基分类器	89.7	0.883
	集成分类器	100	0.995
	SVM分类器	88.3	0.896
6块药	基分类器	90.2	0.915
	集成分类器	99.0	0.995
	SVM分类器	89.3
药块综合	基分类器	92.3
	集成分类器	99.7

表8 故障类型分类的准确率和F1值

Table 8 Accuracy and F1 value statistics for fault classification

故障状态	分类器	准确率/%	F1值
	SVM分类器	89.5	0.872
阻尼失效	基分类器	88.0	0.840
	集成分类器	97.9	0.959
	SVM分类器	91.1	0.863
药仓变形	基分类器	87.2	0.840
	集成分类器	98.0	0.975
	SVM分类器	81.7	0.874
装置健康	基分类器	78.1	0.824
	集成分类器	93.3	0.956
	SVM分类器	87.0
故障综合	基分类器	84.0
	集成分类器	96.3

图12 3个分类器的单项分类准确率

Fig.12 Accuracy of single classification from the three classifiers

参考文献 20

[1]	QIAN H Y, YU Y G, LIU J. Effects of module number and firing condition on charge thermal safety in gun chamber[J]. Defence Technology, 2022, 18(1): 27-37. doi: 10.1016/j.dt.2020.11.003
[2]	WANG X T, XU J, LEI Y, et al. Research on prognostics and health management of tank gun control system[C]// Proceedings of the 2019 International Conference on Sensing, Diagnostics, Prognostics, and Control. Beijing, China: IEEE, 2019: 215-219.
[3]	KANKAR P K, SHARMA S C, HARSHA S P. Fault diagnosis of ball bearings using machine learning methods[J]. Expert Systems with Applications, 2011, 38(3): 1876-1886. doi: 10.1016/j.eswa.2010.07.119 URL
[4]	SOBIE C, FREITAS C, NICOLAI M. Simulation-driven machine learning: bearing fault classification[J]. Mechanical Systems and Signal Processing, 2018, 99: 403-419. doi: 10.1016/j.ymssp.2017.06.025 URL
[5]	PAN X, YU M Y, MENG G L, et al. A study on the diagnosis of compound faults in rolling bearings based on ITD-SVD[J]. Journal of Vibroengineering, 2021, 23(3): 587-602. doi: 10.21595/jve URL
[6]	李良钰, 苏铁熊, 马富康, 等. 基于集合经验模态分解-支持向量机的高压共轨系统故障诊断方法[J]. 兵工学报, 2022, 43(5): 992-1001. doi: 10.12382/bgxb.2021.0155
	LI L Y, SU T X, MA F K, et al. Fault diagnosis method of high-pressure common rail system based on EEMD-SVM[J]. Acta Armamentarii, 2022, 43(5): 992-1001. (in Chinese) doi: 10.12382/bgxb.2021.0155
[7]	HOANG D T, KANG H J. A survey on deep learning based bearing fault diagnosis[J]. Neurocomputing, 2019, 335: 327-335. doi: 10.1016/j.neucom.2018.06.078 URL
[8]	钱林方, 孙乐, 陈光宋, 等. 无位置传感器电机控制在火炮装填应用的关键技术[J]. 兵工学报, 2022, 43(10): 2417-2428.
	QIAN L F, SUN L, CHEN G S, et al. Investigation of the sensorless motor control technology for gun autoloading[J]. Acta Armamentarii, 2022, 43(10): 2417-2428. (in Chinese) doi: 10.12382/bgxb.2022.0330
[9]	刘太素, 钱林方, 陈光宋. 火炮输药机构小射角输药到位一致性稳健优化设计[J]. 兵工学报, 2020, 41(8): 1473-1482. doi: 10.3969/j.issn.1000-1093.2020.08.001
	LIU T S, QIAN L F, CHEN G S. Robust optimal design of a propellant transport mechanism for in-place consistency with small-angle propellant transport[J]. Acta Armamentarii, 2020, 41(8): 1473-1482. (in Chinese) doi: 10.3969/j.issn.1000-1093.2020.08.001
[10]	LEI Y, YANG B, JIANG X W, et al. Applications of machine learning to machine fault diagnosis: a review and roadmap[J]. Mechanical Systems and Signal Processing, 2020, 138: 106587. doi: 10.1016/j.ymssp.2019.106587 URL
[11]	DAI W Y, QIANG Y, XUE G R, et al. Boosting for transfer learning[C]// Proceedings of the 24th International Conference on Machine Learning. New York,NY,US:ACM, 2007: 193-200.
[12]	CHEN W Q, QIU Y, FENG Y H, et al. Diagnosis of wind turbine faults with transfer learning algorithms[J]. Renewable Energy, 2021, 163: 2053-2067. doi: 10.1016/j.renene.2020.10.121 URL
[13]	SHEN F, CHEN C, YAN R Q, et al. Bearing fault diagnosis based on SVD feature extraction and transfer learning classification[C]// Proceedings of 2015 Prognostics and System Health Management Conference. Beijing,China: IEEE, 2015: 1-6.
[14]	XIAO D Y, HUANG Y X, QIN C J, et al. Transfer learning with convolutional neural networks for small sample size problem in machinery fault diagnosis[J]. Proceedings of the Institution of Mechanical Engineers Part C:Journal of Mechanical Engineering Science, 2019, 233(14): 5131-5143. doi: 10.1177/0954406219840381 URL
[15]	HUANG W K, YANG L, GUAN J W. Simulation analysis of transmission system of module propellant storehouse based on ADAMS[C]// Proceedings of the 7th International Conference on Machinery, Material Science and Engineering Application.Hangzhou, China: IOP Publishing, 2021: 012019.
[16]	HU Y C, HOU N N, CHEN C, et al. Interactive feature fusion for end-to-end noise-robust speech recognition[C]// Proceedings of 2022 IEEE International Conference on Acoustics, Speech and Signal Processing.New York,NY,US:IEEE, 2022: 6292-6296.
[17]	LUO Z D, CHEN G. Proper orthogonal decomposition methods for partial differential equations[M]. New York, NY, US: Academic Press, 2018.
[18]	FREUND Y, SCHAPIRE R E. A decision-theoretic generalization of on-line learning and an application to boosting[J]. Journal of Computer and System Sciences, 1997, 55(1): 119-139. doi: 10.1006/jcss.1997.1504 URL
[19]	ZHANG W, DENG L F, ZHANG L, et al. A survey on negative transfer: arXiv:2009.00909[R/OL]. Ithaca, NY, US: Cornell University, 2020(2020-09-02). https://arxiv.org/abs/2009.00909.
[20]	WANG Z R, DAI Z H, PÓCZOS B, et al. Characterizing and avoiding negative transfer[C]// Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition. Long Beach, CA, US:IEEE, 2019: 11293-11302.