Research on diagnosis method for gearbox compound fault using a three-stage hybrid feature selection method
-
摘要: 相比于单部件复合故障，多部件复合故障中故障信息分散在多个域中和特征间相互耦合影响等情况更为严重，导致构造的特征集中往往存在着大量冗余或无关的特征。针对此问题，提出一种三阶段混合式特征选择方法，用于从特征集中筛选敏感特征，提升故障分类准确率。首先，使用4种过滤式模型对故障特征进行评价，然后基于分类错误率对评价结果加权排序，最后结合3种启发式搜索方法按照加权排序结果筛选最优子集。通过一组包含11种故障类别的齿轮–轴承复合故障数据集进行试验，试验结果表明该方法可以在降低特征集维数的同时显著提升分类准确率。Abstract: Unlike single parts compound faults, fault information in multiple parts compound faults is dispersed throughout numerous domains and features are related to one another, resulting in a high number of redundant or irrelevant features in the feature set. A three-stage hybrid feature selection strategy was proposed to overcome this problem by screening sensitive features from feature sets and improving fault classification accuracy. The fault features were first evaluated using four filter models, and the evaluation findings were then weighted and ranked depending on the classification error rate. Finally, based on the weighted sorting results, three heuristic search strategies were merged to select the optimum sub-set. Through a set of gear-bearing compound fault data sets containing 11 fault categories, the result shows that the proposed method may drastically reduce the dimension of the feature set while also boosting classification accuracy.
-
Key words:
- fault diagnosis /
- compound fault /
- feature selection /
- gearbox
-
表 1 特征参数
时域特征参数 频域特征参数 特征值域特征参数 $ {{T}}_{{1}}{=}\dfrac{\displaystyle\sum\nolimits_{n = 1}^{N}{x}\left({n}\right)}{{N}} $ $ {{T}}_{{8}}{=}\dfrac{{{T}}_{{5}}}{{{T}}_{{4}}} $ $ {{F}}_{{1}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{s}\left({k}\right)}{{K}} $ $ {{F}}_{{8}}{=}\sqrt{\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{{{f}}_{{k}}}^{{4}}{s}\left({k}\right)}{\displaystyle\sum\nolimits_{k = 1}^{K}{{{f}}_{{k}}}^{2}{s}\left({k}\right)}} $ $ {{P}}_{{1}}{=}\dfrac{\displaystyle\sum\nolimits_{l = 1}^{N}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)}{{N}} $ $ {{T}}_{{2}}{=}\sqrt{\dfrac{\displaystyle\sum\nolimits_{n = 1}^{N}{\left({x}\left({n}\right)-{{T}}_{{1}}\right)}^{{2}}}{{N}-{1}}} $ $ {{T}}_{{9}}{=}\dfrac{{{T}}_{{5}}}{{{T}}_{{3}}} $ $ {{F}}_{{2}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{\left({s}\left({k}\right)-{{F}}_{{1}}\right)}^{{2}}}{{K}-{1}} $ $ {{F}}_{{9}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{{{f}}_{{k}}}^{2}{s}\left({k}\right)}{\sqrt{\displaystyle\sum\nolimits_{k = 1}^{K}{s}\left({k}\right)\displaystyle\sum\nolimits_{k = 1}^{K}{{{f}}_{{k}}}^{2}{s}\left({k}\right)}} $ $ {{P}}_{{2}}{=}\dfrac{\displaystyle\sum\nolimits_{l = 1}^{N}{{ \lambda }}_{{l}}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)}{\displaystyle\sum\nolimits_{l = 1}^{N}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)} $ $ {{T}}_{{3}}{=}{\left(\dfrac{\displaystyle\sum\nolimits_{n = 1}^{N}\sqrt{\left|{x}\left({n}\right)\right|}}{{N}}\right)}^{{2}} $ $ {{T}}_{{10}}{=}\dfrac{{{T}}_{{4}}}{\dfrac{{1}}{{N}}\displaystyle\sum\nolimits_{n = 1}^{N}\left|{x}\left({n}\right)\right|} $ $ {{F}}_{{3}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{\left({s}\left({k}\right)-{{F}}_{{1}}\right)}^{{3}}}{{K}{\left(\sqrt{{{F}}_{{2}}}\right)}^{{3}}} $ $ {{F}}_{{10}}{=}\dfrac{{{F}}_{{6}}}{{{F}}_{{5}}} $ $ {{P}}_{{3}}{=}\dfrac{\sqrt{\displaystyle\sum\nolimits_{l = 1}^{N}{{(}{{ \lambda }}_{{l}}-{P}_{{2}}{)}}^{{2}}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)}}{\displaystyle\sum\nolimits_{l = 1}^{N}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)} $ $ {{T}}_{{4}}{=}\sqrt{\dfrac{\displaystyle\sum\nolimits_{n = 1}^{N}{\left({x}\left({n}\right)\right)}^{{2}}}{{N}}} $ $ {{T}}_{{11}}{=}\dfrac{{{T}}_{{5}}}{\dfrac{{1}}{{N}}\displaystyle\sum\nolimits_{n = 1}^{N}\left|{x}\left({n}\right)\right|} $ $ {{F}}_{{4}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{\left({s}\left({k}\right)-{{F}}_{{1}}\right)}^{{4}}}{{K}{{{F}}_{{2}}}^{{2}}} $ $ {{F}}_{{11}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{\left({{f}}_{{k}}-{{F}}_{{5}}\right)}^{{3}}{s}\left({k}\right)}{{K}{{{F}}_{{6}}}^{{3}}} $ $ {{P}}_{{4}}{=}\sqrt{\dfrac{\displaystyle\sum\nolimits_{l = 1}^{N}{{{ \lambda }}_{{l}}}^{{2}}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)}{\displaystyle\sum\nolimits_{l = 1}^{N}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)}} $ ${ {T} }_{ {5} }{={\rm{max}}}\left|{x}\left({n}\right)\right|$ $ {{F}}_{{5}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{{f}}_{{k}}{s}\left({k}\right)}{\displaystyle\sum\nolimits_{k = 1}^{K}{s}\left({k}\right)} $ $ {{F}}_{{12}}{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{\left({{f}}_{{k}}-{{F}}_{{5}}\right)}^{{4}}{s}\left({k}\right)}{{K}{{{F}}_{{6}}}^{{4}}} $ $ {{P}}_{{5}}{=}\dfrac{\displaystyle\sum\nolimits_{l = 1}^{N}{{(}{{ \lambda }}_{{l}}-{P}_{{2}}{)}}^{{3}}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)}{{N}{P}_{{3}}^{{4}}} $ $ {{T}}_{{6}}{=}\dfrac{\displaystyle\sum\nolimits_{n = 1}^{N}{{(}{x}\left({n}\right)-{{T}}_{{1}}{)}}^{{3}}}{{(}{N}-{1)}{{{T}}_{{2}}}^{{3}}} $ $ {{F}}_{{6}}{=}\sqrt{\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{\left({{f}}_{{k}}-{{F}}_{{5}}\right)}^{{2}}{s}\left({k}\right)}{{K}}} $ ${ {F} }_{ {13} }{=}\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}\sqrt{\left({ {f} }_{ {k} }-{ {F} }_{ {5} }\right)}{s}\left({k}\right)}{ {K}\sqrt{ { {F} }_{ {6} } } }$ $ {{P}}_{{6}}{=}\dfrac{\displaystyle\sum\nolimits_{l = 1}^{N}{{(}{{ \lambda }}_{{l}}-{P}_{{2}}{)}}^{{4}}\widehat{{f}}\left({{ \lambda }}_{{l}}\right)}{{N}{P}_{{3}}^{{4}}} $ $ {{T}}_{{7}}{=}\dfrac{\displaystyle\sum\nolimits_{n = 1}^{N}{{(}{x}\left({n}\right)-{{T}}_{{1}}{)}}^{{4}}}{{(}{N}-{1)}{{{T}}_{{2}}}^{{4}}} $ $ {{F}}_{{7}}{=}\sqrt{\dfrac{\displaystyle\sum\nolimits_{k = 1}^{K}{{{f}}_{{k}}}^{2}{s}\left({k}\right)}{\displaystyle\sum\nolimits_{k = 1}^{K}{s}\left({k}\right)}} $ 注：表中$ {x}{(}{n}{)} $表示振动信号的时间序列，$ {N} $表示时间序列的采样数目，$ {s}{(}{k}{)} $表示信号$ {x}{(}{n}{)} $的频谱，$ {k=}{1}{,}{2}{,}{\cdots}{,}{K} $，K是谱线数，$ {{f}}_{{k}} $表示第$ {k} $条谱线的频率值，$ \widehat{{f}}\left({{ \lambda }}_{{l}}\right) $表示图信号的特征值谱，$ {{ \lambda }}_{{l}} $表示特征值，$ {l=}{1}{,}{2}{,}{\cdots}{,N} $。 
下载: 导出CSV
表 2 样本类型
故障
类型状态1 状态2 状态3 状态4 状态5 状态6 状态7 状态8 状态9 状态10 状态11 正常
状态√ 齿轮磨
损故障√ √ 齿轮裂
纹故障√ √ 齿轮断
齿故障√ √ √ 轴承内
圈故障√ √ 轴承外圈
故障√ √ 轴承滚动
体故障√ √ √
下载: 导出CSV
表 3 排序结果
排序 FS DET IG PCC 加权 1 227 12 270 269 2 2 221 219 269 270 15 3 222 38 28 263 13 4 12 225 15 264 28 5 1 231 2 258 1 6 13 51 39 257 23 7 219 25 27 106 19 8 217 20 221 228 217 9 118 21 106 222 6 10 120 224 229 233 26 11 119 7 19 234 10 12 228 226 23 251 14 13 121 8 143 252 222 14 131 131 141 2 223 15 14 220 225 15 131
下载: 导出CSV
表 4 实验对比结果
实验 特征个数 识别准确率/(%) 训练时间/s 原始特征集 270 81.7 1.92 实验1 20 93.39 0.96 实验2 20 79.45 0.94 实验3 20 17.21 0.92 本文方法 18 94.42 0.92
下载: 导出CSV
-
[1] Yu F J, Liu Y C, Zhang Q F. Compound fault diagnosis of gearbox based on wavelet packet transform and sparse representation classification[C]. 2020 Chinese Control And Decision Conference (CCDC), 2020: 5339-5343 [2] 周建民, 王发令, 张臣臣, 等. 基于特征优选和GA-SVM的滚动轴承智能评估方法[J]. 振动与冲击, 2021, 40(4): 227-234. [3] Zhang K, Li Y H, Scarf P, et al. Feature selection for high-dimensional machinery fault diagnosis data using multiple models and radial basis function networks[J]. Neurocomputing, 2011, 74(17): 2941-2952. doi: 10.1016/j.neucom.2011.03.043 [4] 胡爱军, 白泽瑞, 赵军. 参数优化VMD结合1.5维谱的滚动轴承复合故障特征分离方法[J]. 振动与冲击, 2020, 39(11): 45-52,62. [5] 张振海, 王维庆, 王海云, 等. 基于HCS-GWO-MSVM的风电机组齿轮箱复合故障诊断研究[J]. 太阳能学报, 2021, 42(10): 176-182. [6] Dhamande L, Chaudhari M. Compound gear-bearing fault feature extraction using statistical features based on time-frequency method[J]. Measurement, 2018, 125: 63-77. doi: 10.1016/j.measurement.2018.04.059 [7] Hu Q, Si X S, Qin A S, et al. Machinery fault diagnosis scheme using redefined dimensionless indicators and mRMR feature selection[J]. IEEE Access, 2020, 8: 40313-40326. doi: 10.1109/ACCESS.2020.2976832 [8] 李帅位, 张栋良, 黄昕宇, 等. 数据特征选择与分类在机械故障诊断中的应用[J]. 振动与冲击, 2020, 39(2): 218-222. doi: 10.13465/j.cnki.jvs.2020.02.030 [9] Zhang X L, Zhang Q, Chen M, et al. A two-stage feature selection and intelligent fault diagnosis method for rotating machinery using hybrid filter and wrapper method[J]. Neurocomputing, 2018, 275: 2426-2439. doi: 10.1016/j.neucom.2017.11.016 [10] Buchaiah S, Shakya P. Bearing fault diagnosis and prognosis using data fusion based feature extraction and feature selection[J]. Measurement, 2022, 188: 110506. doi: 10.1016/j.measurement.2021.110506 [11] Cadenas J, Garrido M, Martinez R. Feature subset selection Filter–Wrapper based on low quality data[J]. Expert Systems with Applications, 2013, 40(16): 6241-6252. doi: 10.1016/j.eswa.2013.05.051 [12] 张润莲, 张昭, 彭小金, 等. 基于Fisher分和支持向量机的特征选择算法[J]. 计算机工程与设计, 2014, 35(12): 4145-4148, 4190. doi: 10.3969/j.issn.1000-7024.2014.12.017 [13] Lei Y G, He Z J, Zi Y Y, et al. Fault diagnosis of rotating machinery based on multiple ANFIS combination with GAs[J]. Mechanical Systems & Signal Processing, 2007, 21(5): 2280-2294. [14] 沈美杰, 赵龙章, 周兵, 等. 基于PSO优化的RBF网络液压泵故障诊断研究[J]. 液压与气动, 2016(5): 87-92. [15] 欧璐, 于德介. 路图傅里叶变换及其在滚动轴承故障诊断中的应用[J]. 机械工程学报, 2015, 51(23): 76-83. -
下载:
点击查看大图
图(6) / 表(4)
计量
- 文章访问数: 174
- HTML全文浏览量: 15
- PDF下载量: 50
- 被引次数: 0
