ADMOW Pattern Recognition Method and its Application on State Recognition of Rolling Bearings
-
摘要: 针对滚动轴承振动信号的状态特征及特征数据中存在"异常值"的现象，提出了基于优化加权代理判别模型（Agent discriminate model based optimization weighted，ADMOW）的模式识别方法。该方法首先通过计算样本特征值的类相似度对特征值进行评价，并依据评价结果对特征值赋予权值，以此弱化"异常值"导致模型出现偏差的问题；然后利用粒子群优化（Particle swarm optimization，PSO）算法对所建立的模型参数进行优化，得到更加准确可靠的分类模型；最后采用建立的优化加权代理预测模型对待测样本进行识别。滚动轴承实验结果表明，与以往的模式识别方法相比，该方法能有效地提高识别准确率。Abstract: Taken the state characteristic of vibration signal of rolling bearing and the "outliers" among the feature values into consideration, a new pattern recognition method, namely agent discriminate model based optimization weighted (ADMOW), is proposed in this paper. Firstly, according to its class similarity, the sample eigenvalues are evaluated. Based on the evaluation result, the eigenvalues are weighted to eliminate the "outliers" resulting in model deviation. The established model is secondly optimized by the particle swarm optimization (PSO) algorithm, hence a more accurate and reliable classification model is correspondingly obtained. Eventually, ADMOW is applied to recognize faults of rolling bearings. The experimental and analytical results indicate that this proposed method can effectively promote the recognition accuracy, compared with the previous pattern recognition methods.
-
Key words:
- ADMOW /
- pattern recognition /
- optimization weighted /
- PSO /
- outliers /
- rolling bearing
-
表 2 ADMOW的特征类可分性评价指标
特征序号 正常状态 内圈故障 外圈故障 滚动体故障 1 0.516 6 0.002 4 0.015 4 0.000 3 2 0.120 6 0.243 6 0.125 4 0.656 9 3 0.239 1 0.498 9 0.469 8 0.068 8 4 0.123 6 0.255 2 0.389 4 0.273 9 
下载: 导出CSV
表 4 各种分类器的分类结果对比
% 分类器 正常轴承 内圈故障 外圈故障 滚动体故障 总识别率 RBF 100 100 74 96 92.5 SVM 94 90 99 100 95.75 VPMCD 98 99 83 97 94.25 加权ADM 100 93 100 100 98.25 优化ADM 100 99 91 100 97.5 ADMOW 100 99 100 100 99.75
下载: 导出CSV
表 5 各种分类器的FRR、FAR指标
% 分类器 性能评价指标 正常轴承 内圈故障 外圈故障 滚动体故障 VPMCD FAR 0 6.67 0 1 FRR 2 1 17 3 加权ADM FAR 0 0 2.33 0 FRR 0 7 0 0 优化ADM FAR 0 3 0 0.33 FRR 0 1 9 0 ADMOW FAR 0 0 0.33 0 FRR 0 1 0 0
下载: 导出CSV
表 6 不同故障尺寸、不同负载下的ADMOW识别率
% 故障尺寸 负载 正常状态 内圈故障 外圈故障 滚动体故障 总识别率 0.01778 0 100 100 100 100 100 0.735 100 100 100 100 100 1.470 100 100 100 100 100 2.205 100 100 100 100 100 0.03556 0 - - - - - 0.735 100 100 100 99 99.75 1.470 100 94 100 100 98.5 2.205 100 96 100 99 98.75 0.05334 0 100 88 93 100 92.75 0.735 100 90 100 100 97.5 1.470 100 100 100 100 100 2.205 100 100 100 99 99.75 注：“-”代表信号数据不足。
下载: 导出CSV
表 7 不同故障尺寸、不同负载下的VPMCD识别率
% 故障尺寸 负载 正常状态 内圈故障 外圈故障 滚动体故障 总识别率 0.01778 0 100 100 97 100 99.25 0.735 100 100 100 100 100 1.470 100 100 100 100 100 2.205 100 100 100 100 100 0.03556 0 - - - - - 0.735 98 52 76 59 71.25 1.470 100 38 83 33 63.5 2.205 100 64 93 57 78.5 0.05334 0 100 84 68 85 84.25 0.735 100 95 88 90 93.25 1.470 100 92 80 81 88.25 2.205 100 99 87 83 92.25
下载: 导出CSV
表 8 各种分类器的分类结果对比
% 分类器 正常轴承 内圈故障 外圈故障 总识别率 VPMCD 100 81.25 100 93.75 ADM 100 82.5.5 100 94.17 加权ADM 100 97.5 98.75 98.75 优化ADM 100 96.25 100 98.75 ADMOW 100 98.75 100 99.58
下载: 导出CSV
-
[1] 韩立群.人工神经网络[M].北京:北京邮电大学出版社, 2006Han L Q. Artificial neural networks[M]. Beijing:Beijing University of Posts and Telecommunications Press, 2006(in Chinese) [2] Wang H Q, Chen P. Intelligent diagnosis method for rolling element bearing faults using possibility theory and neural network[J]. Computers & Industrial Engineering, 2011, 60(4):511-518 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=70ed2b8e6f3d723b2867b41469a8c5dd [3] Fei S W, Zhang X B. Fault diagnosis of power transformer based on support vector machine with genetic algorithm[J]. Expert Systems with Applications, 2009, 36(8):11352-11357 doi: 10.1016/j.eswa.2009.03.022 [4] 袁胜发, 褚福磊.支持向量机及其在机械故障诊断中的应用[J].振动与冲击, 2007, 26(11):29-35, 58 doi: 10.3969/j.issn.1000-3835.2007.11.008Yuan S F, Chu F L. Support vector machines and its applications in machine fault diagnosis[J]. Journal of Vibration and Shock, 2007, 26(11):29-35, 58(in Chinese) doi: 10.3969/j.issn.1000-3835.2007.11.008 [5] Raghuraj R, Lakshminarayanan S. Variable predictive models-a new multivariate classification approach for pattern recognition applications[J]. Pattern Recognition, 2009, 42(1):7-16 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=f0d3234dfabd36150b51f8c84c5a0bf5 [6] Raghuraj R, Lakshminarayanan S. Variable predictive model based classification algorithm for effective separation of protein structural classes[J]. Computational Biology and Chemistry, 2008, 32(4):302-306 doi: 10.1016/j.compbiolchem.2008.03.009 [7] Cressie N. The origins of kriging[J]. Mathematical Geology, 1990, 22(3):239-252 doi: 10.1007/BF00889887 [8] Gou P, Liu W, Cui W C. A comparison of approximation methods for multidisciplinary design optimization of ship structures[J]. Journal of Ship Mechanics, 2007, 11(6):913-923 http://www.wanfangdata.com.cn/details/detail.do?_type=perio&id=cblx200706012 [9] 金欣磊.基于PSO的多目标优化算法研究及应用[D].杭州: 浙江大学, 2006Jin X L. PSO-based multi-objective optimization algorithm research and its applications[D]. Hangzhou: Zhejiang University, 2006(in Chinese) [10] 隋文涛, 路长厚, 张丹.基于加权FCM算法的轴承故障诊断[J].武汉理工大学学报, 2010, 34(1):72-75 http://d.old.wanfangdata.com.cn/Periodical/whjtkjdxxb201001018Sui W T, Lu C H, Zhang D. Bearings fault diagnosis based on feature weighted FCM algorithm[J]. Journal of Wuhan University of Technology, 2010, 34(1):72-75(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/whjtkjdxxb201001018 [11] 唐俊.PSO算法原理及应用[J].计算机技术与发展, 2010, 20(2):213-216 doi: 10.3969/j.issn.1673-629X.2010.02.056Tang J. Principle and application of PSO algorithm[J]. Computer Technology and Development, 2010, 20(2):213-216(in Chinese) doi: 10.3969/j.issn.1673-629X.2010.02.056 [12] 李良敏, 任景岩.基于粒子群算法优化的独立分量分析算法[J].振动与冲击, 2015, 34(8):7-11, 25 http://d.old.wanfangdata.com.cn/Periodical/zdycj201508003Li L M, Ren J Y. A universal particle swarm-optimized independent component analysis algorithm[J]. Journal of Vibration and Shock, 2015, 34(8):7-11, 25(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/zdycj201508003 [13] 潘海洋, 杨宇, 郑近德, 等.基于径向基函数的变量预测模型模式识别方法[J].航空动力学报, 2017, 32(2):500-506. http://d.old.wanfangdata.com.cn/Periodical/hkdlxb201702031Pan H Y, Yang Y, Zheng J D, et al. Variable predictive model based RBF class discriminate method[J]. Journal of Aerospace Power, 2017, 32(2):500-506(in Chinese) http://d.old.wanfangdata.com.cn/Periodical/hkdlxb201702031 [14] 程军圣, 郑近德, 杨宇.一种新的非平稳信号分析方法-局部特征尺度分解法[J].振动工程学报, 2012, 25(2):215-220 doi: 10.3969/j.issn.1004-4523.2012.02.017Cheng J S, Zheng J D, Yang Y. A nonstationary signal analysis approach-the local characteristic-scale decomposition method[J]. Journal of Vibration Engineering, 2012, 25(2):215-220(in Chinese) doi: 10.3969/j.issn.1004-4523.2012.02.017 [15] 毕雪芹, 苏艳娟, 王琪.嵌入式指纹识别系统的设计及试验研究[J].国外电子测量技术, 2015, 34(2):50-53, 81 doi: 10.3969/j.issn.1002-8978.2015.02.011Bi X Q, Su Y J, Wang Q. Design and experimental study of embedded fingerprint identification system[J]. Foreign Electronic Measurement Technology, 2015, 34(2):50-53, 81(in Chinese) doi: 10.3969/j.issn.1002-8978.2015.02.011 -
点击查看大图
图(8) / 表(8)
计量
- 文章访问数: 306
- HTML全文浏览量: 166
- PDF下载量: 17
- 被引次数: 0
