Multi-objective fusion diagnosis of aeroengine wear failure
-
摘要:
针对多种油液分析数据的特点，建立了航空发动机磨损故障融合诊断方法，实现基于油液分析数据的航空发动机磨损状态综合评估。该故障融合诊断方法包括磨损故障定性分析、定位分析和定因分析。定性分析以光谱、铁谱和颗粒计数原始分析数据为输入，基于（Dempster-Shafer）证据理论获得发动机磨损故障定性诊断结果；在定位分析部分，建立了基于深度学习的滚动轴承故障部位识别模型，以能谱分析原始数据作为模型输入，实现了航空发动机磨损部位的智能识别；最后，在定性分析部分，利用定性结果和定位结果，根据领域专家的经验，建立了基于if-then的知识规则，找出发动机磨损故障原因；利用实际油液监测数据对所提方法的有效性和可靠性进行验证，诊断精度最高可达到100%，结果充分表明了该方法的正确性、有效性。
Abstract:According to the characteristics of various oil analysis data, an aeroengine wear fault fusion diagnosis method was established to realize comprehensive evaluation of aeroengine wear state based on oil analysis data. The fault fusion diagnosis method included wear fault qualitative analysis, location analysis and cause analysis. Taking the original analysis data of spectrum, Ferrography and particle count as the input, the qualitative diagnosis results of engine wear fault were obtained based on D-S evidence theory through qualitative analysis; in the location analysis, a rolling bearing fault location identification model based on deep learning was established, and the original data of energy spectrum analysis were used as the model input to realize the intelligent identification of aeroengine wear location; finally, in the cause analysis, using the qualitative results and positioning results, according to the experience of domain experts, the knowledge rules based on if-then were established to find out the cause of engine wear fault. The effectiveness and reliability of the proposed method were verified by using the actual oil monitoring data, the diagnostic accuracy can reach up to 100%, and the results fully showed the correctness and effectiveness of the method.
-
图 1 基于多种油液分析数据的航空发动机磨损故障多目标融合诊断流程图
Figure 1. Multi-objective fusion diagnosis of aeroengine wear fault based on multiple oil analysis data oil analysis flow chart
图 3 一维卷积神经网络训练流程图
Figure 3. One-dimensional convolutional neural network training flow chart
表 1 各种油液分析数据对故障诊断的有效性分析
Table 1. Effectiveness analysis of various oil analysis data for fault diagnosis
参数 光谱数据 颗粒计数数据 铁谱数据 能谱数据 理化数据 定性 *** *** *** — * 定位 * — — *** — 定因 — ** *** — *** 注：“***”表示有效性最高；“**”表示有效性中等；“*”表示有效性较差；“—”为无效。 
下载: 导出CSV
表 2 规则可信度
Table 2. Rule credibility
证据 规则可信度 故障分析方法 Fe 0.5 光谱分析 Cr 0.05 Pb 0.05 Cu 0.1 Sn 0.05 Al 0.05 Ni 0.02 Ti 0.02 Mn 0.05 Ag 0.02 Si 0.05 Mg 0.02 Mo 0.02 疲劳磨粒 0.4 铁谱分析 球状磨粒 0.1 层状磨粒 0.1 红色氧化物 0.2 黑色氧化物 0.2 >5 μm 0.1 颗粒计数分析 >15 μm 0.2 >25 μm 0.3 >50 μm 0.4
下载: 导出CSV
表 3 一维卷积残差网络参数
Table 3. One dimensional convolution residual network parameters
结构 卷积核参数 输出大小 第1层 （1×3×64）×2 1×31 第2层 （1×3×128）×2 1×15×128 第3层 （1×3×256）×2 1×7×256 第4层 （1×3×512）×2 1×3×512 FC 512×3×1 1536 ×1LSTM 1536 ×3×200200×1 FC 全连接层 29×1
下载: 导出CSV
表 4 油液性能参数变化与故障原因对照表
Table 4. Comparison between changes of oil performance parameters and fault causes
油液分析类型 油液参数变化 故障发生原因 铁谱分析 疲劳、球状和层状磨粒数量高于警告限 滚动轴承疲劳剥落失效 红色氧化物数量高于警告限 润滑油中混有水分 黑色氧化物数量高于警告限 润滑油供应不足，长时间高温高负载工作 理化分析 水分含量高于警告限 润滑油中混有水分 杂质含量高于警告限 润滑油中混有硬质颗粒杂质 黏度含量高于警告限 润滑油供应不足，长时间高温高负载工作 黏度含量低于正常限 润滑油中混有水分 酸值含量高于警告限 润滑油供应不足
下载: 导出CSV
表 5 元素故障界限值
Table 5. Element fault limit value
属性 铁谱分析/（个/mL） 理化分析 疲劳
磨粒球状
磨粒层状
磨粒红色
氧化物黑色
氧化物黏度/
（mm2/s）酸度/
（mg/kg）闪点
t/℃水分/
（mg/kg）杂质/
%正常值 2 3 1 1 1 25 0.05 258 0 0 警告值 4 5 2 3 2 22.5/27.5 0.15 273 0.001 0.1
下载: 导出CSV
表 6 用于定性诊断的油液数据（部分数据）
Table 6. Fluid data for qualitative diagnosis （partial data）
数据 光谱分析/‰ 铁谱分析/（个/mL） 颗粒计数分析/（个/100 mL） Fe Ag Cu Cr 疲劳
磨粒球状
磨粒层状
磨粒红色
氧化物>5 μm >15 μm >25 μm >50 μm 1 1.054 0.056 0.62 0.36 2.6 3.45 0.64 2.05 2501.255 181.213 92.736 8.931 2 6.039 0.033 4.33 0.12 3 3.3 0.62 2.23 1799.37 522.769 290.427 11.405 3 1.162 0.079 0.932 0.21 2.9 3.4 0.53 2.13 2011.551 42.114 89.304 10.713 4 0.582 0.059 1.313 0 3 3.36 0.64 1.69 5630.736 213.183 203.268 13.845 5 0.674 0.081 2.063 0.27 2.7 3.38 0.60 1.52 3112.509 88.515 91.591 15.163 6 1.008 0.094 2.316 0.3 3.6 3.93 0.66 2.1 3358.563 101.445 109.875 19.181 7 0.799 0.056 2.605 0.29 3.1 3.85 1.22 2.1 3512.164 233.834 114.392 22.239 40 6.674 0.203 9.99 0.29 7.4 5.2 1.95 2.4 8088.109 1221.251 396.159 51.003
下载: 导出CSV
表 7 用于定位诊断的油液数据（部分数据）
Table 7. Fluid data for location diagnosis （partial data）
‰ 数据 Cu Zn Al Mn Fe Sn Cr Mo V O C 1 0 0 0 0 89.99 0 4.10 4.80 1.11 0 0 2 0 0 0 0 89.63 0 4.25 4.78 1.35 0 0 3 0 0 0 0 93.81 0 3.54 2.11 0.54 0 0 4 61.98 31.17 3.00 1.84 1.37 0.64 0 0 0 0 0 5 0 0 0 0 69.21 0 0 0 0 30.79 0 6 0 0 0 0 66.45 0 0 0 0 28.67 4.88 7 0 0 63.97 0 17.63 0 0 0 0 4.73 13.67 40 0 0 0 0 94.04 0 0 0.14 0.11 0 5.71
下载: 导出CSV
表 8 用于定因诊断的油液数据（部分数据）
Table 8. Fluid data for cause determination diagnosis （partial data）
数据 铁谱分析/（个/mL） 理化分析 疲劳
磨粒球状
磨粒层状
磨粒红色
氧化物黑色
氧化物黏度/
（mm2/s）酸度/
（mg/kg）闪点
t/℃水分/
（mg/kg）杂质/% 1 2.6 3.45 0.64 2.05 0.17 24.87 0.05 256.37 0 0.03 2 3 3.3 0.62 2.23 0 24.83 0.04 256.72 0 0.01 3 2.9 3.4 0.53 2.13 0.28 25.13 0.05 257.3 0 0.1 4 3 3.36 0.64 1.69 0.04 25.01 0.07 258.82 0 0.1 5 2.7 3.38 0.60 1.52 0.03 25.14 0.05 257.78 0 0.12 6 3.6 3.93 0.66 2.1 0.09 25.17 0.05 258.49 0 0.11 7 3.1 3.85 1.22 2.1 0.26 25.12 0.05 259.92 0 0.12 40 7.4 5.2 1.95 2.4 0.95 25.62 0.05 269.43 0 0.155
下载: 导出CSV
表 9 光谱诊断结果（y=0.661）
Table 9. Spectral diagnosis results （y=0.661）
证据 A B C Fe 1 0.5 0.5 Ag 0.0517 0.02 0.001 Cu 1 0.1 0.1 Cr 0.49 0.05 0.245
下载: 导出CSV
表 10 铁谱诊断结果（y=0.743）
Table 10. Ferrographic diagnosis results （y=0.743）
证据 A B C 疲劳磨粒 1 0.4 0.5 球状磨粒 0.78 0.1 0.312 层状磨粒 0.65 0.1 0.065 红色氧化物 0.72 0.2 0.144 黑色氧化物 0.33 0.2 0.066
下载: 导出CSV
表 11 颗粒计数诊断结果（y=0.696）
Table 11. Particle count diagnostic results （y=0.696）
证据 A B C >5 μm 1 0.1 0.1 >15 μm 0.98 0.2 0.196 >25 μm 1 0.3 0.3 >50 μm 1 0.4 0.4
下载: 导出CSV
表 12 光谱和铁谱融合诊断结果（y=0.912）
Table 12. Fusion diagnosis results of spectrum and ferrographic （y=0.912）
证据 A B C Fe 1 0.5 0.5 Ag 0.0517 0.02 0.001 Cu 1 0.1 0.1 Cr 0.49 0.05 0.245 疲劳磨粒 1 0.4 0.5 球状磨粒 0.78 0.1 0.312 层状磨粒 0.65 0.1 0.065 红色氧化物 0.72 0.2 0.144 黑色氧化物 0.33 0.2 0.066
下载: 导出CSV
表 13 定性融合诊断结果（y=0.973）
Table 13. Qualitative fusion diagnosis results （y=0.973）
证据 A B C Fe 1 0.5 0.5 Ag 0.0517 0.02 0.001 Cu 1 0.1 0.1 Cr 0.49 0.05 0.245 疲劳磨粒 1 0.4 0.5 球状磨粒 0.78 0.1 0.312 层状磨粒 0.65 0.1 0.065 红色氧化物 0.72 0.2 0.144 黑色氧化物 0.33 0.2 0.066 >5 μm 1 0.1 0.1 >15 μm 0.98 0.2 0.196 >25 μm 1 0.3 0.3 >50 μm 1 0.4 0.4
下载: 导出CSV
表 14 定位诊断结果
Table 14. Location diagnosis results
序号 Resnet18 Resnet34 LSTM CNN 本文模型 1 0Cr18Ni9 ZL114A-T6 1Cr13 1Cr17Ni2 1Cr11Ni2W2MoV 84% 93% 90% 82% 95% 2 镍石墨 1Cr11Ni2W2MoV 2Cr13 40CrNiMoA 2Cr13 77% 89% 86% 79% 91% 3 1Cr12Ni2WMoVNb 18Cr2Ni4WA ZG1Cr18Ni9Ti 1Cr13 1Cr13 77% 76% 84% 66% 89% 4 1Cr17Ni2 1Cr17Ni2 1Cr11Ni2W2MoV 1Cr18Ni9Ti 40CrNiMoA 72% 74% 79% 64% 85% 5 1Cr11Ni2W2MoV 1Cr12Ni2WMoVNb 1Cr17Ni2 2Cr13 1Cr17Ni2 65% 71% 73% 53% 80%
下载: 导出CSV
表 15 定因诊断结果
Table 15. Causal diagnosis results
元素值 是否处于
异常状态疲劳磨粒数量>4 个/mL 是 球状磨粒数量>5 个/mL 是 层状磨粒数量<2 个/mL 否 红色氧化物数量<3 个/mL 否 黑色氧化物数量<2 个/mL 否 22.2 mm2/s<黏度<27.5 mm2/s 否 酸值<0.15 mg/kg 否 闪点值t<273 ℃ 否 水分含量<0.001 mg/kg 否 杂质含量<0.1% 否
下载: 导出CSV
-
[1] 庞梦洋,索中英,郑万泽,等. 基于RS-CART决策树的航空发动机小样本故障诊断[J]. 航空动力学报,2020,35(7): 1559-1568. PANG Mengyang,SUO Zhongying,ZHENG Wanze,et al. Small sample fault diagnosis of aeroengine based on RS-CART decision tree[J]. Journal of Aerospace Power,2020,35(7): 1559-1568. (in ChinesePANG Mengyang, SUO Zhongying, ZHENG Wanze, et al. Small sample fault diagnosis of aeroengine based on RS-CART decision tree[J]. Journal of Aerospace Power, 2020, 35(7): 1559-1568. (in Chinese) [2] ZHAO Yongping,WANG Jianjun,LI Xiaoya,et al. Extended least squares support vector machine with applications to fault diagnosis of aircraft engine[J]. ISA Transactions,2020,97: 189-201. doi: 10.1016/j.isatra.2019.08.036 [3] DUAN Zhihe,WU Tonghai,GUO Shuaiwei,et al. Development and trend of condition monitoring and fault diagnosis of multi-sensors information fusion for rolling bearings: a review[J]. The International Journal of Advanced Manufacturing Technology,2018,96(1): 803-819. [4] MURALI S,XIA Xingao,JAGTIANI A V,et al. Capacitive coulter counting: detection of metal wear particles in lubricant using a microfluidic device[J]. Smart Materials and Structures,2009,18(3): 037001. doi: 10.1088/0964-1726/18/3/037001 [5] 王洪伟,陈果,陈立波,等. 一种航空发动机滚动轴承磨损故障监测技术[J]. 航空动力学报,2014,29(9): 2256-2263. WANG Hongwei,CHEN Guo,CHEN Libo,et al. A fault monitoring technique for wear of aero-engine rolling bearing[J]. Journal of Aerospace Power,2014,29(9): 2256-2263. (in ChineseWANG Hongwei, CHEN Guo, CHEN Libo, et al. A fault monitoring technique for wear of aero-engine rolling bearing[J]. Journal of Aerospace Power, 2014, 29(9): 2256-2263. (in Chinese) [6] 张赟,林学森,王琳,等. 采用监督局部切空间排列算法的航空发动机磨损故障诊断[J]. 西安交通大学学报,2020,54(4): 179-185. ZHANG Yun,LIN Xuesen,WANG Lin,et al. Aero-engine wear fault diagnosis with supervised locally tangent space alignment[J]. Journal of Xi’an Jiaotong University,2020,54(4): 179-185. (in ChineseZHANG Yun, LIN Xuesen, WANG Lin, et al. Aero-engine wear fault diagnosis with supervised locally tangent space alignment[J]. Journal of Xi’an Jiaotong University, 2020, 54(4): 179-185. (in Chinese) [7] 陈果. 基于神经网络和D-S证据理论的发动机磨损故障融合诊断[J]. 航空动力学报,2005,20(2): 303-308. CHEN Guo. Fusion diagnosis of engine wearing fault based on neural networks and D-S evidence theory[J]. Journal of Aerospace Power,2005,20(2): 303-308. (in Chinese doi: 10.3969/j.issn.1000-8055.2005.02.026CHEN Guo. Fusion diagnosis of engine wearing fault based on neural networks and D-S evidence theory[J]. Journal of Aerospace Power, 2005, 20(2): 303-308. (in Chinese) doi: 10.3969/j.issn.1000-8055.2005.02.026 [8] 姜旭峰,宗营,阮少军. 基于光谱-铁谱分析的航空发动机磨损故障诊断应用研究[J]. 化工时刊,2020,34(4): 1-5. JIANG Xufeng,ZONG Ying,RUAN Shaojun. ApplicationResearch of wear fault diagnosis of aero-engine based on spectrum-ferrography analysis[J]. Chemical Industry Times,2020,34(4): 1-5. (in ChineseJIANG Xufeng, ZONG Ying, RUAN Shaojun. ApplicationResearch of wear fault diagnosis of aero-engine based on spectrum-ferrography analysis[J]. Chemical Industry Times, 2020, 34(4): 1-5. (in Chinese) [9] 孔祥兴,潘燕,武通海. 考虑多监测指标不确定性的滑油状态综合评估方法[J]. 航空发动机,2020,46(4): 71-77. KONG Xiangxing,PAN Yan,WU Tonghai. Comprehensive assessment method of oil status considering uncertainty of multiple monitoring index[J]. Aeroengine,2020,46(4): 71-77. (in ChineseKONG Xiangxing, PAN Yan, WU Tonghai. Comprehensive assessment method of oil status considering uncertainty of multiple monitoring index[J]. Aeroengine, 2020, 46(4): 71-77. (in Chinese) [10] 张鹏飞,李本威,韦祥,等. 基于滑油监控信息的发动机磨损状态融合评估[J]. 航空发动机,2015,41(4): 93-97. ZHANG Pengfei,LI Benwei,WEI Xiang,et al. Evaluation of aeroengine wear condition fusion based on oil monitoring information[J]. Aeroengine,2015,41(4): 93-97. (in ChineseZHANG Pengfei, LI Benwei, WEI Xiang, et al. Evaluation of aeroengine wear condition fusion based on oil monitoring information[J]. Aeroengine, 2015, 41(4): 93-97. (in Chinese) [11] 於迪,贺石中,何伟楚,等. 基于油液监测的立磨磨辊故障诊断分析[J]. 润滑油,2019,34(3): 45-49,54. YU Di,HE Shizhong,HE Weichu,et al. Fault diagnosis analysis of vertical grinding roller based on oil monitoring[J]. Lubricating Oil,2019,34(3): 45-49,54. (in ChineseYU Di, HE Shizhong, HE Weichu, et al. Fault diagnosis analysis of vertical grinding roller based on oil monitoring[J]. Lubricating Oil, 2019, 34(3): 45-49, 54. (in Chinese) [12] 马安祥,李艳军,曹愈远,等. 基于免疫理论的航空发动机磨损故障智能诊断[J]. 航空学报,2015,36(6): 1896-1904. MA Anxiang,LI Yanjun,CAO Yuyuan,et al. Intelligent diagnosis for aircraft engine wear fault based on immune theory[J]. Acta Aeronautica et Astronautica Sinica,2015,36(6): 1896-1904. (in ChineseMA Anxiang, LI Yanjun, CAO Yuyuan, et al. Intelligent diagnosis for aircraft engine wear fault based on immune theory[J]. Acta Aeronautica et Astronautica Sinica, 2015, 36(6): 1896-1904. (in Chinese) [13] 张海涛. 基于在线油液监测的船舶发动机故障分析[J]. 机械设计与制造工程,2020,49(3): 113-116. ZHANG Haitao. Fault analysis on the marine engine based on online oil monitoring[J]. Machine Design and Manufacturing Engineering,2020,49(3): 113-116. (in Chinese doi: 10.3969/j.issn.2095-509X.2020.03.024ZHANG Haitao. Fault analysis on the marine engine based on online oil monitoring[J]. Machine Design and Manufacturing Engineering, 2020, 49(3): 113-116. (in Chinese) doi: 10.3969/j.issn.2095-509X.2020.03.024 [14] BASIR O,YUAN X. Engine fault diagnosis based on multi-sensor information fusion using Dempster-Shafer evidence theory[J]. Information Fusion,2007,8(4): 379-386. [15] 陈果,陈立波,宋兰琪. 一种磨损故障融合诊断新方法及其应用[J]. 机械科学与技术,2009,28(9): 1157-1161. CHEN Guo,CHEN Libo,SONG Lanqi. A new approach to fusion diagnosis of wear faults and its application[J]. Mechanical Science and Technology for Aerospace Engineering,2009,28(9): 1157-1161. (in Chinese doi: 10.3321/j.issn:1003-8728.2009.09.009CHEN Guo, CHEN Libo, SONG Lanqi. A new approach to fusion diagnosis of wear faults and its application[J]. Mechanical Science and Technology for Aerospace Engineering, 2009, 28(9): 1157-1161. (in Chinese) doi: 10.3321/j.issn:1003-8728.2009.09.009 [16] 王古常,成坚,鲍传美,等. 模糊推理和证据理论融合的航空发动机故障诊断[J]. 航空动力学报,2011,26(9): 2101-2106. WANG Guchang,CHENG Jian,BAO Chuanmei,et al. Diagnosis method of aero-engine fault based on fuzzy inference and evidence theory[J]. Journal of Aerospace Power,2011,26(9): 2101-2106. (in ChineseWANG Guchang, CHENG Jian, BAO Chuanmei, et al. Diagnosis method of aero-engine fault based on fuzzy inference and evidence theory[J]. Journal of Aerospace Power, 2011, 26(9): 2101-2106. (in Chinese) [17] ZHUANG Zilong,LV Huichun,XU Jie,et al. A deep learning method for bearing fault diagnosis through stacked residual dilated convolutions[J]. Applied Sciences,2019,9(9): 1823. doi: 10.3390/app9091823 [18] WANG Youming,CHENG Lin. A combination of residual and long-short-term memory networks for bearing fault diagnosis based on time-series model analysis[J]. Measurement Science and Technology,2021,32(1): 015904. doi: 10.1088/1361-6501/abaa1e [19] 张西宁,向宙,唐春华. 一种深度卷积自编码网络及其在滚动轴承故障诊断中的应用[J]. 西安交通大学学报,2018,52(7): 1-8,59. ZHANG Xining,XIANG Zhou,TANG Chunhua. A deep convolutional auto-encoding neural network and its application in bearing fault diagnosis[J]. Journal of Xi’an Jiaotong University,2018,52(7): 1-8,59. (in Chinese doi: 10.7652/xjtuxb201807001ZHANG Xining, XIANG Zhou, TANG Chunhua. A deep convolutional auto-encoding neural network and its application in bearing fault diagnosis[J]. Journal of Xi’an Jiaotong University, 2018, 52(7): 1-8, 59. (in Chinese) doi: 10.7652/xjtuxb201807001 -
点击查看大图
计量
- 文章访问数: 135
- HTML浏览量: 61
- PDF量: 30
- 被引次数: 0