A method for identifying bearing lubricating oil multi-debris based on two-level neural network
-
摘要:
提出了基于误差反向传播神经网络的滑油多屑末识别方法，解决了多屑末信号重叠的难题。提出两级模型框架，第一级网络能准确估计重叠信号中微小屑末的数量，第二级网络利用数量信息，精准分析微小屑末的直径，成功克服了信号重叠带来的挑战。通过充分数据训练和模型结构优化，模型在单屑末、双屑末和三屑末识别上取得了98.10%、91.42%和93.06%准确率。
Abstract:An innovative method for identifying multiple debris in lubricating oil based on back propagation neural networks was introduced to address the challenge of multi-debris signal overlap. A two-level model framework was proposed, of which the first level network can accurately estimate the number of small debris in overlapping signals, and the second-level network utilized this quantity information to precisely analyze the diameter of these small debris, successfully overcoming the challenges posed by signal overlap. Through sufficient data training and model structure optimization, the model achieved 98.10%, 91.42%, and 93.06% accuracy, respectively, in single, double, and triple debris recognition.
-
图 6 双屑末信号与屑末间隔关系图
Figure 6. Diagram of the relationship between dual debris signal and debris intervals
图 7 双屑末信号与屑末尺寸关系图
Figure 7. Diagram of the relationship between dual debris signal and debris size
图 9 不同结构两级网络正确率对比
Figure 9. Comparison of accuracy of two level networks with different structures
图 11 单屑末数量误判与直径关系
Figure 11. Relationship between misjudgment of single debris quantity and diameter
图 13 三屑末误判与最小间隔关系
Figure 13. Relationship between misjudgment and the minimum interval of three debris
图 14 不同位置屑末直径判断误差分布图
Figure 14. Distribution chart of diameter determination errors for different positional debris
图 15 多屑末直径误判与屑末直径比关系图
Figure 15. Relationship chart between misjudgment of multi-debris diameter and debris diameter ratio
表 1 相邻屑末间隔分布
Table 1. Distribution of adjacent debris intervals
屑末间隔 3～6 mm 7～12 mm 13～17 mm 18 mm以上 百分比/% 25.11 26.85 23.79 24.25 
下载: 导出CSV
表 2 相邻屑末直径比例分布
Table 2. Distribution of diameter ratios of adjacent debris
直径比例 1～1.33 1.33～1.66 1.66～2 2以上 百分比/% 58.84 26.73 11.29 3.14
下载: 导出CSV
表 3 各网络结构参数
Table 3. Parameters of each network structure
参数 数值或说明 屑末
数量识别单屑末
直径判断双屑末
直径判断三屑末
直径判断隐藏层 3层 1层 2层 3层 每层节点 120 80 100 100 正确率/% 97.20 100 94.62 95.45
下载: 导出CSV
表 4 屑末数量判断网络正确率
Table 4. Accuracy of network judgment based on number of debris
% 项目 单屑末 双屑末 三屑末 正确率 98.10 96.62 97.50 误判为单屑末 1.38 0 误判为双屑末 1.90 2.50 误判为三屑末 0 2.00 综合正确率 97.30
下载: 导出CSV
表 5 屑末直径判断模型正确率
Table 5. Accuracy of debris diameter judgment model
% 项目 单屑末序号 双屑末序号 三屑末序号 1 1 2 1 2 3 最大误差 8.1 26.28 25.03 27.02 38.52 16.95 平均误差 1.02 3.13 2.89 2.34 2.56 1.82 中位误差 0.65 1.93 1.70 1.66 1.57 1.36 正确率 100 94.62 95.45
下载: 导出CSV
表 6 两级网络判断模型综合正确率
Table 6. Comprehensive accuracy of two-level network judgment model
% 项目 单屑末 双屑末 三屑末 数量判断正确率 98.10 96.62 97.5 尺寸判断正确率 100 94.62 95.45 综合正确率 98.10 91.42 93.06
下载: 导出CSV
-
[1] 马佳丽,陈果,康玉祥,等. 航空发动机磨损故障多目标融合诊断[J]. 航空动力学报,2024,39(10): 20220191. MA Jiali,CHEN Guo,KANG Yuxiang,et al. Multi-objective fusion diagnosis of aeroengine wear fault[J]. Journal of Aerospace Power,2024,39(10): 20220191. (in ChineseMA Jiali, CHEN Guo, KANG Yuxiang, et al. Multi-objective fusion diagnosis of aeroengine wear fault[J]. Journal of Aerospace Power, 2024, 39(10): 20220191. (in Chinese) [2] 曹明,黄金泉,周健,等. 民用航空发动机故障诊断与健康管理现状、挑战与机遇Ⅰ: 气路、机械和FADEC系统故障诊断与预测[J]. 航空学报,2022,43(9): 625573. CAO Ming,HUANG Jinquan,ZHOU Jian,et al. Current status,challenges and opportunities of civil aero-engine diagnostics & health management Ⅰ: Diagnosis and prognosis of engine gas path,mechanical and FADEC[J]. Acta Aeronautica et Astronautica Sinica,2022,43(9): 625573. (in ChineseCAO Ming, HUANG Jinquan, ZHOU Jian, et al. Current status, challenges and opportunities of civil aero-engine diagnostics & health management Ⅰ: Diagnosis and prognosis of engine gas path, mechanical and FADEC[J]. Acta Aeronautica et Astronautica Sinica, 2022, 43(9): 625573. (in Chinese) [3] 万洋,魏麟,刘晔璐,等. 一种航空发动机智能检屑装置设计[J]. 仪表技术与传感器,2019(6): 23-28. WAN Yang,WEI Lin,LIU Yelu,et al. Design of aeroengine intelligent debris detection device[J]. Instrument Technique and Sensor,2019(6): 23-28. (in Chinese doi: 10.3969/j.issn.1002-1841.2019.06.006WAN Yang, WEI Lin, LIU Yelu, et al. Design of aeroengine intelligent debris detection device[J]. Instrument Technique and Sensor, 2019(6): 23-28. (in Chinese) doi: 10.3969/j.issn.1002-1841.2019.06.006 [4] 陆凤霞,韦坤,王春雷,等. 直升机中减三相流金属屑末可达性[J]. 航空学报,2024,45(1): 128524. LU Fengxia,WEI Kun,WANG Chunlei,et al. Accessibility of metal particles in three-phase flow of helicopter intermediate gearbox[J]. Acta Aeronautica et Astronautica Sinica,2024,45(1): 128524. (in ChineseLU Fengxia, WEI Kun, WANG Chunlei, et al. Accessibility of metal particles in three-phase flow of helicopter intermediate gearbox[J]. Acta Aeronautica et Astronautica Sinica, 2024, 45(1): 128524. (in Chinese) [5] WU Yu,ZHANG Hongpeng. Solid particles,water drops and air bubbles detection in lubricating oil using microfluidic inductance and capacitance measurements[J]. Journal of Micromechanics and Microengineering,2019,29(2): 025011. doi: 10.1088/1361-6439/aaf7f0 [6] 杨昊,孙衍山,李健,等. 滑油磨粒信号的变分模态分解和概率密度估计[J]. 仪器仪表学报,2018,39(4): 99-106. YANG Hao,SUN Yanshan,LI Jian,et al. Variational mode decomposition and probability density estimation of lubricating oil debris detection signal[J]. Chinese Journal of Scientific Instrument,2018,39(4): 99-106. (in ChineseYANG Hao, SUN Yanshan, LI Jian, et al. Variational mode decomposition and probability density estimation of lubricating oil debris detection signal[J]. Chinese Journal of Scientific Instrument, 2018, 39(4): 99-106. (in Chinese) [7] MOURA E E,BERBERT P A,BERBERT-MOLINA M A,et al. Performance analysis of RF dielectric models for density-independent estimation of moisture content in sorghum[J]. Powder Technology,2013,246: 369-378. doi: 10.1016/j.powtec.2013.04.030 [8] REN Yi,ZHAO Guo,QIAN Min,et al. A highly sensitive triple-coil inductive debris sensor based on an effective unbalance compensation circuit[J]. Measurement Science and Technology,2019,30(1): 015108. doi: 10.1088/1361-6501/aaf119 [9] QIAN Min,REN Yijun,FENG Zhihua. Wear debris sensor using intermittent excitation for high sensitivity,wide detectable size range,and low heat generation[J]. IEEE Transactions on Industrial Electronics,2023,70(6): 6386-6394. doi: 10.1109/TIE.2022.3190894 [10] WESER R,WÖCKEL S,HEMPEL U,et al. Particle characterization in highly concentrated suspensions by ultrasound scattering method[J]. Sensors and Actuators A: Physical,2013,202: 30-36. doi: 10.1016/j.sna.2013.02.027 [11] WU Tonghai,PENG Yeping,WU Hongkun,et al. Full-life dynamic identification of wear state based on on-line wear debris image features[J]. Mechanical Systems and Signal Processing,2014,42(1/2): 404-414. [12] 曾周末,许恩蕾,黄新敬,等. 高灵敏低压电磁感应式滑油磨屑传感器[J]. 仪器仪表学报,2022,43(2): 1-9. ZENG Zhoumo,XU Enlei,HUANG Xinjing,et al. High-sensitivity,low-voltage lubricating oil wear debris sensor based on electromagnetic induction[J]. Chinese Journal of Scientific Instrument,2022,43(2): 1-9. (in ChineseZENG Zhoumo, XU Enlei, HUANG Xinjing, et al. High-sensitivity, low-voltage lubricating oil wear debris sensor based on electromagnetic induction[J]. Chinese Journal of Scientific Instrument, 2022, 43(2): 1-9. (in Chinese) [13] FLANAGAN I M,JORDAN J R,WHITTINGTON H W. An inductive method for estimating the composition and size of metal particles[J]. Measurement Science and Technology,1990,1(5): 381-384. doi: 10.1088/0957-0233/1/5/001 [14] MILLER J L,KITALJEVICH D. In-line oil debris monitor for aircraft engine condition assessment[R]. Piscataway,US: 2000 IEEE Aerospace Conference,2000. [15] REN Yijun,LI Wei,ZHAO Guofeng,et al. Inductive debris sensor using one energizing coil with multiple sensing coils for sensitivity improvement and high throughput[J]. Tribology International,2018,128: 96-103. doi: 10.1016/j.triboint.2018.07.025 [16] 贾宝惠,丁梓航,肖海建,等. 基于梳状式电容传感器的滑油磨粒检测方法[J]. 仪器仪表学报,2022,43(12): 49-57. JIA Baohui,DING Zihang,XIAO Haijian,et al. Detection method of lube wear particle based on comb type capacitance sensor[J]. Chinese Journal of Scientific Instrument,2022,43(12): 49-57. (in ChineseJIA Baohui, DING Zihang, XIAO Haijian, et al. Detection method of lube wear particle based on comb type capacitance sensor[J]. Chinese Journal of Scientific Instrument, 2022, 43(12): 49-57. (in Chinese) [17] 孙衍山,杨昊,佟海滨,等. 航空发动机滑油磨粒在线监测[J]. 仪器仪表学报,2017,38(7): 1561-1569. SUN Yanshan,YANG Hao,TONG Haibin,et al. Review of on-line detection for wear particles in lubricating oil of aviation engine[J]. Chinese Journal of Scientific Instrument,2017,38(7): 1561-1569. (in Chinese doi: 10.3969/j.issn.0254-3087.2017.07.001SUN Yanshan, YANG Hao, TONG Haibin, et al. Review of on-line detection for wear particles in lubricating oil of aviation engine[J]. Chinese Journal of Scientific Instrument, 2017, 38(7): 1561-1569. (in Chinese) doi: 10.3969/j.issn.0254-3087.2017.07.001 [18] HONG Wei,CAI Wenjian,WANG Shaoping,et al. Mechanical wear debris feature,detection,and diagnosis: a review[J]. Chinese Journal of Aeronautics,2018,31(5): 867-882. doi: 10.1016/j.cja.2017.11.016 [19] XIAO Hong,WANG Xinyu,LI Hongcheng,et al. An inductive debris sensor for a large-diameter lubricating oil circuit based on a high-gradient magnetic field[J]. Applied Sciences,2019,9: 1546. doi: 10.3390/app9081546 [20] REN Yijun,WEN Xiaolong,GAO Shengwei,et al. Theoretical and simulation analysis on the spatial resolution of magnetic metal debris sensors[J]. Frontiers in Materials,2023,10: 1110834. doi: 10.3389/fmats.2023.1110834 [21] WU Sen,LIU Zhijian,YUAN Haichao,et al. Multichannel inductive sensor based on phase division multiplexing for wear debris detection[J]. Micromachines,2019,10(4): 246. doi: 10.3390/mi10040246 [22] CHEN Tao,WANG Liyong,GU Yuhai,et al. Review on online inductive wear debris monitoring technology[J]. The Journal of Engineering,2019,2019(23): 8518-8521. doi: 10.1049/joe.2018.9045 [23] WU Xianwei,LIU Hairui,QIAN Zhi,et al. On the investigation of frequency characteristics of a novel inductive debris sensor[J]. Micromachines,2023,14(3): 669. doi: 10.3390/mi14030669 [24] WU Xianwei,ZHANG Yinghong,LI Nian,et al. A new inductive debris sensor based on dual-excitation coils and dual-sensing coils for online debris monitoring[J]. Sensors,2021,21(22): 7556. doi: 10.3390/s21227556 -
点击查看大图
计量
- 文章访问数: 283
- HTML浏览量: 154
- PDF量: 30
- 被引次数: 0