Identification Method of Shaft Orbit in Rotating Machines using EHF-TCDs and SVM
-
摘要: 针对运用轴心轨迹进行旋转机械故障诊断时,存在提取特征困难和识别率低等问题,在精确型高度函数EHF1(Extract height function 1)和TCDs(Triangular centroid cistances)描述子的基础上提出了一种EHF-TCDs描述子,并使用平滑化和傅里叶变换对其进行降维,该描述子具有起始点不变性、相似变换不变性、抗噪性、低维度等特点,并能充分表征轴心轨迹,再使用SVM对提取的EHF-TCDs描述子特征进行训练与测试,进而提出了一种新的旋转机械故障快速诊断方法。通过一个无噪声和4个有噪声的模拟轴心轨迹库和一个实测轴心轨迹库验证了该方法的有效性,其识别率都在99.57%以上,单个样本平均测试时间不超过0.021 ms。Abstract: In order to overcome the difficulty in extracting the suitable features of the shaft orbit and low efficiency of the identification on shaft orbit in fault detection for rotating machines, a shape descriptor, EHF-TCDs, based on Extract height function 1 (EHF1) and Triangular centroid distances (TCDs) is presented to extract feature from the shaft orbit of rotating machines. Smoothing and Fourier transforms are introduced into the shape descriptor to reduce the dimension of the feature matrix. EHF-TCDs shape descriptor has the characteristics of starting point invariance, similarity transformation invariance, noise immunity and low dimension, and can fully characterize shaft orbits. On the basis of EHF-TCDs, a new and efficient method of fault detection for rotating machines is proposed, which uses support vector machine (SVM) to identify the EHF-TCDs feature extracted from shaft orbit. The effectiveness of the proposed method is verified by a noiseless simulated shaft orbit library, four noisy simulated shaft orbit libraries and a measured shaft orbit library, and the recognition rates of experiments all exceeded 99.57%, and the average test time of a single sample does not exceed 0.021 ms.
-
Key words:
- rotating machines /
- shaft orbits /
- fault detection /
- feature extraction /
- support vector machines /
- shape descriptor /
- identification
-
表 1 不同描述子特征矩阵的维数比较
描述子 特征维数 SC N×Nd×Nθ =100×8×12=9 600 IDSC N×Nd×Nθ =100×8×12=9 600 EHF1 N×H =100×20=2 000 TCDs (N-1)×H =99×16=1 584 EHF-TCDs 2×M×H =2×16×16=512 表 2 旋转机械故障对应的轴心轨迹图形
故障类型 轴心轨迹形状 转子不对中 香蕉型 转子不平衡 椭圆形 油膜涡动 内8字 转子不对中 外8字 油膜震荡 花瓣形 表 3 各描述子提取特征时间
ms 描述子 单个无噪声样本提取特征时间/ms 单个有噪声样本提取特征时间/ms EHF-TCDs 17.985 18.075 EHF1 18.016 18.214 TCDs 16.187 16.239 IDSC 46.426 46.624 SC 23.063 23.062 表 4 模拟无噪声轴心轨迹实验结果
算法 平均识别率/% 特征提取后的训练时间/s 单个样本平均测试时间/ms 本文提出算法 99.95 0.011 0.015 SVM+SC 98.02 3.456 7.648 SVM+IDSC 96.46 3.592 7.631 SVM+TCDs 96.04 0.407 0.457 SVM+EHF1 95.81 0.563 0.778 BP+EHF-TCDs 99.44 2.761 0.041 BP+EHF1 96.50 70.16 0.924 BP+TCDs 94.10 57.17 0.726 BP+SC 87.84 368.6 3.044 BP+IDSC 83.38 364.8 3.013 表 5 不同信噪比的模拟有噪声轴心轨迹识别率
% 算法 不同信噪比的识别率/% 5 dB 10 dB 15 dB 20 dB 本文提出算法 99.74 99.80 99.91 99.93 SVM+SC 93.20 95.72 97.06 97.18 SVM+TCDs 91.76 93.66 94.73 95.87 SVM+IDSC 90.48 92.64 93.38 95.02 SVM+EHF1 87.14 89.46 92.06 93.66 BP+EHF-TCDs 98.53 99.06 99.20 99.38 BP+EHF1 90.19 92.75 93.65 93.77 BP+TCDs 88.53 90.20 91.67 92.22 BP+SC 81.00 81.09 86.60 87.18 BP+IDSC 72.21 75.25 80.89 82.62 -
[1] 孙慧芳, 潘罗平, 张飞, 等.旋转机械轴心轨迹识别方法综述[J].中国水利水电科学研究院学报, 2014, 12(1):86-92 http://d.old.wanfangdata.com.cn/Periodical/zgslsdkxyjyxb201401015Sun H F, Pan L P, Zhang F, et al. Review of identification of shaft orbit for rotating machinery[J]. Journal of China Institute of Water Resources and Hydropower Research, 2014, 12(1):86-92(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/zgslsdkxyjyxb201401015 [2] 耿富礼, 李富才, 孟光.二维全息谱在旋转机械故障诊断中的应用[J].机械科学与技术, 2014, 33(10):1445-1449 http://www.cnki.com.cn/Article/CJFDTotal-JXKX201410002.htmGeng F L, Li F C, Meng G. Research on fault diagnosis using 2-D holospectrum for rotating machinery[J]. Mechanical Science and Technology for Aerospace Engineering, 2014, 33(10):1445-1449(in Chinese) http://www.cnki.com.cn/Article/CJFDTotal-JXKX201410002.htm [3] 孙国栋, 艾成汉, 周振, 等.基于高度函数的旋转机械轴心轨迹识别方法[J].中国测试, 2017, 43(9):118-122 http://d.old.wanfangdata.com.cn/Periodical/zgcsjs201709021Sun G D, Ai C H, Zhou Z, et al. Axis orbit identification of rotating machine based on height function[J]. China Measurement & Testing Technology, 2017, 43(9):118-122(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/zgcsjs201709021 [4] 陈坚, 叶渊杰, 陈抒, 等.基于不变矩和神经网络的泵机组轴心轨迹自动识别[J].排灌机械工程学报, 2011, 29(1):67-71 doi: 10.3969/j.issn.1674-8530.2011.01.014Chen J, Ye Y J, Chen S, et al. Automatic identification of pump unit axis orbit based on invariant moments features and neural networks[J]. Journal of Drainage and Irrigation Machinery Engineering, 2011, 29(1):67-71(in Chinese) doi: 10.3969/j.issn.1674-8530.2011.01.014 [5] Zhou J Z, Xiao H, Li C S, et al. Shaft orbit identifica-tion for rotating machinery based on statistical fuzzy vector chain code and support vector machine[J]. Journal of Vibroengineering, 2014, 16(2):713-724 [6] Bao J T, Zhu Z W, Tang H R, et al. Apply low-level image feature representation and classification method to identifying shaft orbit of hydropower unit[C]//Proceedings of 2014 Sixth International Conference on Intelligent Human-Machine Systems and Cybernetics. Hangzhou, China: IEEE, 2014: 165-168 [7] 陈晓玥, 周建中, 肖剑, 等.一种模仿人眼的汽轮机轴心轨迹识别方法[J].振动、测试与诊断, 2015, 35(4):677-684 http://d.old.wanfangdata.com.cn/Periodical/zdcsyzd201504013Chen X Y, Zhou J Z, Xiao J, et al. A shaft orbit identification method imitating human eyes for steam turbine[J]. Journal of Vibration, Measurement & Diagnosis, 2015, 35(4):677-684(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/zdcsyzd201504013 [8] Chen X Y, Zhou J Z, Xiao H, et al. Fault diagnosis based on comprehensive geometric characteristic and probability neural network[J]. Applied Mathematics and Computation, 2014, 230:542-554 doi: 10.1016/j.amc.2013.12.122 [9] Belongie S, Malik J, Puzicha J. Shape matching and object recognition using shape contexts[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2002, 24(4):509-522 doi: 10.1109/34.993558 [10] Ling H B, Jacobs D W. Shape classification using the inner-distance[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2007, 29(2):286-299 doi: 10.1109/TPAMI.2007.41 [11] 孙国栋, 张杨, 李萍, 等.用于快速形状匹配的精确型高度函数特征描述[J].光学精密工程, 2017, 25(1):224-235 http://d.old.wanfangdata.com.cn/Periodical/gxjmgc201701029Sun G D, Zhang Y, Li P, et al. Feature description of exact height function used in fast shape retrieval[J]. Optics and Precision Engineering, 2017, 25(1):224-235(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/gxjmgc201701029 [12] Yang C Z, Wei H, Yu Q, et al. A novel method for 2D nonrigid partial shape matching[J]. Neurocomputing, 2018, 275:1160-1176 doi: 10.1016/j.neucom.2017.09.067 [13] 徐乐, 邢邦圣, 郎超男, 等.LMD能量熵和SVM相结合的滚动轴承故障诊断[J].机械科学与技术, 2017, 36(6):915-918 http://www.cnki.com.cn/Article/CJFDTOTAL-JXKX201706015.htmXu L, Xing B S, Lang C N, et al. Fault diagnosis of rolling bearing combined LMD energy entropy and SVM[J]. Mechanical Science and Technology for Aerospace Engineering, 2017, 36(6):915-918(in Chinese) http://www.cnki.com.cn/Article/CJFDTOTAL-JXKX201706015.htm [14] 张蕾, 周洲.基于小波和信息粒化的BP神经网络的轴承故障诊断[J].机械科学与技术, 2012, 31(1):49-52 doi: 10.3969/j.issn.1672-1616.2012.01.013Zhang L, Zhou Z. The BP neural network fault diagnosis of bearings based on wavelet and information granulation[J]. Mechanical Science and Technology for Aerospace Engineering, 2012, 31(1):49-52(in Chinese) doi: 10.3969/j.issn.1672-1616.2012.01.013 [15] Wang J W, Bai X, You X E, et al. Shape matching and classification using height functions[J]. Pattern Recognition Letters, 2012, 33(2):134-143 http://d.old.wanfangdata.com.cn/NSTLQK/NSTL_QKJJ0225879282/ [16] 王军伟.融合全局与局部信息的形状轮廓特征分析与匹配[D].武汉: 华中科技大学, 2012Wang J W. Research on shape contour analysis and matching based on combined global and local information[D]. Wuhan: Huazhong University of Science and Technology, 2012(in Chinese) [17] Chang C C, Lin C J. LIBSVM:a library for support vector machines[J]. ACM Transactions on Intelligent Systems and Technology, 2011, 2(3):27 http://d.old.wanfangdata.com.cn/Periodical/jdq201315008