基于故障敏感度学习和集成学习的可诊断性设计方法

doi:10.12382/bgxb.2023.0413

摘要/Abstract

摘要：

对系统进行可诊断性设计,能优化诊断方案、提升诊断方案与系统的契合程度,对于准确检测或隔离系统故障具有重大意义。为此提出一种基于故障敏感度学习和集成学习的可诊断性设计方法。对于采集到的不同状态下的系统测试点信号进行特征提取,结合系统的可诊断性评价结果,提出特征贡献度量化算法,对信号中不同特征对故障诊断的贡献度进行评估,使用改进的D-S证据理论算法对不同信号的特征进行融合,确定适用于故障检测与隔离的故障敏感特征集合;采用集成学习方法对基分类器的诊断效果进行加强,最终获得对当前系统的诊断方案。仿真实验结果表明,采用新的可诊断性设计方法设计出来的针对系统的诊断方案,可以对系统故障进行良好的诊断,其中与未进行故障敏感度学习环节相比,故障诊断的错误率由5.33%下降到2.66%,与未进行集成学习环节相比,故障诊断的错误率由16.22%(基诊断器的平均值)下降到2.66%。在与其他诊断方案的对比实验中,新方法的故障诊断错误率相较于对比方法的平均值下降了3.34%。

关键词: 可诊断性设计, 故障敏感度, 集成学习, 自适应提升

Abstract:

The diagnosability design of system can optimize the diagnostic scheme and improve the the degree of compatibility between diagnostic scheme and system, which is of great significance for the accurate detection or isolation of system faults. For this, a diagnosability design method based on fault-sensitive learning and integrated learning is proposed. The featuresfrom the collected signals of system test points in different states are extracted. Combined with the results of diagnosability evaluation of the system, a quantitative feature contribution algorithm is proposed to evaluate the contribution degree of different features in the signal to fault diagnosis. The improved D-S evidence theory algorithm is used to fuse the features of different signals, which can determine the fault sensitive feature set suitable for fault detection and isolation. The integrated learning method is used to strengthen the diagnostic effect of base classifier, and finally the diagnosis scheme of the current system is obtained. Simulation experiments show that the system diagnosis scheme designed by the proposed diagnosability design method can make a good diagnosis of system faults. Compared with no fault-sensitive learning, the error rate of fault diagnosis decreases from 5.33% to 2.66%. Compared with no integrated learning, the error rate of fault diagnosis decreases from 16.22% (mean value of base diagnostics) to 2.66%. In the comparison experiments with other diagnostic schemes, the fault diagnosis error rate of the proposed method is decreased by 3.34% compared to the average fault diagnosis error rateof other methods.

Key words: diagnosability design, fault sensitivity, integratedlearning, adaptive boosting

中图分类号:

TP306⁺.3

吕佳朋, 史贤俊, 王元鑫. 基于故障敏感度学习和集成学习的可诊断性设计方法[J]. 兵工学报, 2024, 45(7): 2451-2462.

LÜ Jiapeng, SHI Xianjun, WANG Yuanxin. Diagnosability Design Method Based on Fault-sensitive Learning and Integrated Learning[J]. Acta Armamentarii, 2024, 45(7): 2451-2462.

图/表 19

图1 基于知识的故障诊断方法原理示意图

Fig.1 Schematic diagram of knowledge-based fault diagnosis method

图2 故障可诊断性设计框架

Fig.2 Fault diagnosability design diagram

图3 故障敏感度学习方法流程图

Fig.3 Flow chart of fault-sensitive learning method

图4 OVA分类示意图

Fig.4 Diagram of OVA architecture classification

图5 分类器集成学习示意图

Fig.5 Schematic diagram of classifier integrated learning

图6 电路示意图

Fig.6 Circuit diagram

表1 电路故障设置

Table 1 Circuit fault setting

故障	元件	故障值
f₀	-
f₁	C₁/nf	10
f₂	R_s/kΩ	50
f₃	R_1s/kΩ	短路
f₄	R_1s/kΩ	20

表2 信号特征种类

Table 2 Types of signal characteristics

序号	信号特征	表达式
1	最大值	S_max= $m a x i$ $\mathscr{S}$ (i)
2	最小值	S_min= $m i n i$ $\mathscr{S}$ (i)
3	平均值	S_ave= $1 N ∑ i = 1 N$ $\mathscr{S}$ (i)
4	均方根	S_RMS= $1 N ∑ i = 1 N S (i) 2$
5	峰峰值	S_peak=S_max-S_min
6	峰值因子	S_peakF=S_peak/S_rms
7	峭度	S_kurtosis= $1 N ∑ i = 1 N (S (i) - S a v e) 4 [1 N ∑ i = 1 N (S (i) - s a v e) 2] 2$
8	偏度	S_skewness= $1 N ∑ i = 1 N (S (i) - S a v e) 4 [1 N ∑ i = 1 N (S (i) - s a v e) 2] 32$
9	波形因子	S_wave_F=S_RMS/( $1 N ∑ i = 1 N$ \| $\mathscr{S}$ (i)\|)
10	脉冲因子	S_impulseF=S_peak/( $1 N ∑ i = 1 N$ \| $\mathscr{S}$ (i)\|)
11	裕度因子	S_clearF=S_peak/( $1 N$ ( $∑ i = 1 N \| S (i) \|$ )²)

表2 信号特征种类

Table 2 Types of signal characteristics

序号	信号特征	表达式
1	最大值	S_max= $m a x i$ $\mathscr{S}$ (i)
2	最小值	S_min= $m i n i$ $\mathscr{S}$ (i)
3	平均值	S_ave= $1 N ∑ i = 1 N$ $\mathscr{S}$ (i)
4	均方根	S_RMS= $1 N ∑ i = 1 N S (i) 2$
5	峰峰值	S_peak=S_max-S_min
6	峰值因子	S_peakF=S_peak/S_rms
7	峭度	S_kurtosis= $1 N ∑ i = 1 N (S (i) - S a v e) 4 [1 N ∑ i = 1 N (S (i) - s a v e) 2] 2$
8	偏度	S_skewness= $1 N ∑ i = 1 N (S (i) - S a v e) 4 [1 N ∑ i = 1 N (S (i) - s a v e) 2] 32$
9	波形因子	S_wave_F=S_RMS/( $1 N ∑ i = 1 N$ \| $\mathscr{S}$ (i)\|)
10	脉冲因子	S_impulseF=S_peak/( $1 N ∑ i = 1 N$ \| $\mathscr{S}$ (i)\|)
11	裕度因子	S_clearF=S_peak/( $1 N$ ( $∑ i = 1 N \| S (i) \|$ )²)

表3 故障可检测性评价结果

Table 3 Fault detectability evaluation results

故障可检测性和测点	故障
故障可检测性和测点	f₁	f₂	f₃	f₄
故障可检测性	0.4207	0.3120	0.1871	0.5884
测点	测点3	测点3	测点6	测点6

表4 故障可隔离性评价结果

Table 4 Fault isolability evaluation results

故障	f₁	f₂	f₃	f₄
f₁	0	0.7327	0.4207	0.9664
f₂	测点3	0	0.4679	0.3204
f₃	测点3	测点6	0	0.7751
f₄	测点6	测点6	测点6	0

图7 信号特征对故障诊断贡献度

Fig.7 Contribution of signal characteristics to fault diagnosis

表5 诊断算法诊断结果

Table 5 Diagnosis results of diagnostic algorithm

故障及合计	f₀	f₁	f₂	f₃	f₄	合计
诊断错误率/%	0	6.67	0	6.67	0	2.66

表6 不同诊断器的诊断结果

Table 6 Diagnostic results of different diagnosers %

故障及合计	基分类器1	基分类器2	基分类器3	基分类器4	基分类器5	基分类器6	集成分类器
f₀	0	13.33	26.67	0	6.67	80	0
f₁	0	0	26.67	6.67	13.33	100	6.67
f₂	6.67	13.33	33.33	0	0	66.67	0
f₃	6.67	6.67	13.33	0	6.67	13.33	6.67
f₄	0	0	20	6.67	6.67	13.33	0
合计	2.66	6.67	24	2.66	6.67	54.7	2.66

图8 基诊断器及集成诊断器诊断效果对比

Fig.8 Comparison of diagnostic effects of base diagnoser and integrated diagnose

表7 不同特征组合对于集成诊断结果的影响

Table 7 Effects of different feature combinations on integrated diagnosis results %

序号	特征组合	f₀	f₁	f₂	f₃	f₄	合计
1	1 2 3 4 5 6	0	6.67	0	6.67	0	2.66
2	1 2	40	6.67	20	13.33	0	16
3	4 5 6	0	0	6.67	6.67	6.67	4
4	4 5	13.33	6.67	26.67	20	6.67	14.67
5	4	20	0	6.67	13.33	6.67	9.33
6	5	53.3	33.33	53.3	53.3	20	42.67
7	1 2 4 5	40	6.67	20	20	0	17.33
8	1 2 3 4 5 6 7 8 9 10 11	6.67	0	6.67	13.33	0	5.33
9	7 8	46.67	20	80	80	0	45.33
10	9 10 11	66.67	33.33	73.33	53.3	6.67	46.67
11	7 8 9 10 11	33.33	26.67	66.67	73.33	6.67	41.33

表8 不同诊断方案诊断结果

Table 8 Diagnostic results of different diagnosis schemes %

方法	f₁	f₂	f₃	f₄	合计
本文方法	6.67	0	6.67	0	2.66
文献[25]方法	0	26	13.33	0	8
文献[26]方法	0	0	13.33	6.67	4

图9 集成学习环节多次实验数据分布对比

Fig.9 Comparison of data distributions of multiple experiments in integrated learning sessions

图10 3种方法经多次实验的数据分布对比

Fig.10 Comparison of the data distributions of three methods after several experiments

图11 3种方法在不同训练与测试样本比下的诊断结果

Fig.11 Diagnostic results of three methods with different training to test sample ratios

参考文献 26

[1]	LIU Y K, GUO L, GAO H L, et al. Machine vision based condition monitoring and fault diagnosis of machine tools using information from machined surface texture: a review[J]. Mechanical Systems and Signal Processing, 2022, 164(1):108068.
[2]	GUO J C, SI Z T, XIANG J W. A compound fault diagnosis method of rolling bearing based on wavelet scattering transform and improved soft threshold denoising algorithm[J]. Measurement, 2022, 196(15):111276.
[3]	MA Y, SHI H B, TAN S, et al. Double branch rebalanced network for class imbalance learning in process fault diagnosis[J]. Industrial & Engineering Chemistry Research, 2023, 62(2):1057-1072.
[4]	YANG S J, YANG P K, YU H, et al. A2DCNN-RF model for offshore wind turbine high-speed bearing-fault diagnosis under noisy environment[J]. Energies, 2022, 15(9):1-16.
[5]	李昕燃, 靳伍银. 基于改进麻雀算法优化支持向量机的滚动轴承故障诊断研究[J]. 振动与冲击, 2023, 42(6):106-114.
	LI X R, JIN W Y. Fault diagnosis of rolling bearings based on ISSA-SVM[J]. Journal of Vibration and Shock, 2023, 42(6):106-114. (in Chinese)
[6]	张刘润, 郑近德, 潘海洋, 等. 基于全矢autogram的滚动轴承故障诊断方法[J]. 振动测试与诊断, 2023, 43(2): 312-320,410-411.
	ZHANG L R, ZHENG J D, PAN H Y, et al. Full vector autogram based fault diagnosis method for rolling bearing[J]. Journal of Vibration and Shock, 2023, 43(02): 312-320,410-411. (in Chinese)
[7]	王天舒, 余刃, 毛伟, 等. 基于知识矩阵的核动力装置运行故障诊断方法[J]. 海军工程大学学报, 2022, 34(6):73-78,83.
	WANG T S, YU R, MAO W, et al. Fault diagnosis algorithm of nuclear power plant based on knowledge matrix[J]. Journal of Naval University of Engineering, 2022, 34(6):73-78,83. (in Chinese)
[8]	邵庆祝, 谢民, 王同文, 等. 基于深度随机配置网络的电网故障诊断方法[J]. 控制工程, 2022, 29(12): 2213-2220.
	SHAO Q Z, XIE M, WANG T W, et al. Diagnosis method of power grid fault based on deep stochastic configuration network[J]. Control Engineering of China, 2022, 29(12):2213-2220. (in Chinese)
[9]	许水清, 黄文展, 何怡刚, 等. 基于自适应滑模观测器的中点钳位型三电平并网逆变器开路故障诊断[J]. 电工技术学报, 2023, 38(4):1010-1022.
	XU S Q, HUANG W Z, HE Y G, et al. Open-circuit fault diagnosis method of neutral point clamped three-level grid-connected inverter based on adaptive sliding mode observer[J]. Transactions of China Electrotechnical Society, 2023, 38(4):1010-1022. (in Chinese)
[10]	GAO Z W, CECATI C, DING S X. A survey of fault diagnosis and fault-tolerant techniques—Part Ⅰ:fault diagnosis with model-based and signal-based approaches[J]. IEEE Transactions on Industrial Electronics, 2015, 62(6):3757-3767.
[11]	GAO Z W, CECATI C, DING S X. A survey of fault diagnosis and fault-tolerant techniques-Part II:fault diagnosis with knowledge-based and hybrid/active approaches[J]. IEEE Transactions on Industrial Electronics, 2015, 62(6):3768-3774.
[12]	姜午恺, 周平. 非平衡数据下基于自适应迁移与宽度学习的多工况故障诊断[J]. 中国科学:信息科学, 2023, 53(5): 946-971.
	JIANG W K, ZHOU P. Multi-condition fault diagnosis based on adaptive transfer and broad learning under imbalanced data[J]. Scientia SinicaInformationis, 2023, 53(5):946-971. (in Chinese)
[13]	QIN Y F, SHI X J. Fault diagnosis method for rolling bearings based on two-channel CNN under unbalanced datasets[J]. Applied Sciences, 2022, 12(17):8474.
[14]	李泽东, 李志农, 陶俊勇, 等. 基于特征融合的注意力增强卷积神经网络的航空发动机滚动轴承故障诊断方法[J]. 兵工学报, 2022, 43(12):3228-3239.
	LI Z D, LI Z N, TAO J Y, et al. Fault diagnosis for aero-engine rolling bearings based on an attention augmented convolutional neural network with feature fusion[J]. Acta Armamentarii, 2022, 43(12):3228-3239. (in Chinese)
[15]	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
[16]	朱良玉, 崔倩文, 胡超凡, 等. 基于子结构最优传输的跨工况轴承故障诊断方法[J]. 振动与冲击, 2023, 42(7): 273-280,332.
	ZHU L Y, CUI Q W, HU C F, et al. Fault diagnosis method of bearing under cross working conditions based on substructure optimal transmission[J]. Journal of Vibration and Shock, 2023, 42(7):273-280,332. (in Chinese)
[17]	周长巍, 李国勇, 任密蜂, 等. 基于区分性联合概率分布的域适应故障诊断[J]. 振动与冲击, 2023, 42(7): 170-179.
	ZHOU C W, LI G Y, REN M F, et al. Domain adaption fault diagnosis based on discriminative joint probability distribution difference[J]. Journal of Vibration and Shock, 2023, 42(7):170-179. (in Chinese)
[18]	管业勤, 蒋祖华. 基于elman网络和ecoc-svm的机电系统并发多故障诊断方法[J]. 哈尔滨工程大学学报, 2020, 41(11):1715-1720.
	GUAN Y Q, JIANG Z H. A concurrent multifault diagnosis method for electromechanical systems based on the Elman network and an ECOC-SVM[J]. Journal of Harbin Engineering University, 2020, 41(11): 1715-1720. (in Chinese)
[19]	王大轶, 符方舟, 刘成瑞, 等. 控制系统可诊断性的内涵与研究综述[J]. 自动化学报, 2018, 44(9): 1537-1553.
	WANG D Y, FU F Z, LIU C R, et al. Connotation and research status of diagnosability of control systems: areview[J]. Acta Automatica Sinica, 2018, 44(9): 1537-1553. (in Chinese)
[20]	WANG D Y, FU F Z, LI W B, et al. A review of diagnosability of control systems with applications to spacecraft[J]. Annual Reviews in Control, 2020, 49: 212-229.
[21]	SHI X J, QIN Y F, ZHAO L. Optimal test point placement based on fault diagnosability quantitative evaluation[J]. IEEE Access, 2022, 10:74495-74507.
[22]	李萌萌, 刘艺, 李庚松, 等. 不平衡多分类算法综述[J]. 计算机应用, 2022, 42(11):3307-3321.
	LI M M, LIU Y, LI G S, et al. Survey on imbalanced multi-class classification algorithms[J]. Journal of Computer Application, 2022, 42(11):3307-3321. (in Chinese)
[23]	贾邵华. 基于冷启动svdd技术的模拟电路模块级故障诊断方法研究[D]. 哈尔滨: 哈尔滨工业大学, 2020.
	JIA S H. Research on fault diagnosis of analog circuit modules based on cold start SVDD theropy[D]. Harbin: Harbin Institute of Technology, 2020. (in Chinese)
[24]	吕佳朋, 史贤俊, 聂新华, 等. 基于递归图和张量分解的故障可诊断性评价方法[J]. 兵工学报, 2023, 44(3):763-772. doi: 10.12382/bgxb.2021.0790
	LÜ J P, SHI X J, NIE X H, et al. Evaluationmehod of fault diagnosability based on recurisive plot and tensor decomposition[J]. Acta Armamentarii, 2023, 44(3):763-772. (in Chinese)
[25]	王雨虹, 孙远星, 包伟川, 等. 基于数据均衡化与改进鲸鱼算法优化核极限学习机的变压器故障诊断方法[J]. 信息与控制, 2023, 52(2):235-244. doi: 10.13976/j.cnki.xk.2023.2112
	WANG Y H, SUN Y X, BAO W C, et al. Transformer fault diagnosis method based on data equalization and kernel-based extreme learning machine of improved whale algorithm[J]. Information and Control, 2023, 52(2): 235-244. (in Chinese) doi: 10.13976/j.cnki.xk.2023.2112
[26]	陈冠宇, 杨鹏, 陈宁. 基于随机森林算法的船舶电站故障诊断[J]. 船舶工程, 2023, 45(1):116-119.
	CHEN G Y, YANG P, CHEN N. Fault diagnosis of ship power station based on random forest algorithm[J]. Ship Engineering, 2023, 45(1):116-119. (in Chinese)