Classification of Unbalanced Fault Data Based on Similarity Factor Analysis of Sliding Window
-
摘要: 针对机械装备监测过程中不均衡故障数据难以辨识,提出了一种基于滑动窗口相似性因子分析方法。该方法引入滑动窗口技术,通过分析目标数据与历史数据的PCA相似性因子,从旧的过程数据中筛选出与诊断目标相似的数据,构成待选数据池;然后采用距离相似性因子,从待选数据池中选择出与目标数据最相似的数据用于辅助训练。将该方法用于转子故障的不均衡数据分类中,在不同偏斜率下采用KPCA-SVM方法进行故障分类。结果表明:该方法可有效地改善分类决策边界,降低由样本不均衡而引起的误诊断率。Abstract: Aiming at difficult identification of unbalanced fault data for the mechanical equipment monitoring, anovel method based on similarity factor analysis of sliding mode control is proposed.In this method, the slidingmode control technology is introduced, and a dataset pool is constituted by analyzing the principal componentanalysis similarity factor between the target data and historical date, which is from the old monitoring dataset.Thenthe most similar data which will be used in training is obtained by analyzing the distance similarity factor of targetdata and the data from the dataset pool.The kernel principal component analysis-support vector machines method isapplied in the unbalanced data classification of fault rotor.The result shows that this method can improve theclassification decision boundary effectively, and reduce the false diagnosis caused by the unbalanced samples.
-
Key words:
- data transfer /
- efficiency /
- eigenvalues and eigenfunctions /
- failure analysis
-
[1] 于达仁,胡清华,鲍文.融合粗糙集合模糊聚类的连续数据知识发现[J].中国电机工程学报,2004,24(6):205-210Yu D R,Hu Q H,Bao W.Combining rough setmethodology and fuzzy clustering for knowledge discoveryfrom quantitative data[J].Proceedings of the CSEE,2004,24(6): 205-210 (in Chinese) [2] Scholkopf B,Smola A,Muller K.Nonlinear componentanalysis as a kernel eigenvalue problem[J].NeuralComputation,1998,10(5): 1299-1319 [3] 陶新民,童智靖,刘玉,等.基于ODR 和BSMOTE 结合的不均衡数据SVM 算法[J].控制与决策,2011,26(10): 1535-1541Tao X M,Tong Z J,Liu Y,et al.SVM classifier forunbalanced data based on combination of ODR andBSMOTE[J].Control and Decision,2011,26 (10):1535-1541 (in Chinese) [4] 唐明珠,王岳斌,阳春华.一种改进的支持向量数据描述故障诊断方法[J].控制与决策,2011,26 (7):967-972Tang M Z,Wang Y B,Yang C H.Modified supportvector data description for fault diagnosis[J].Controland Decision,2011,26(7): 967-972 (in Chinese) [5] 刘天羽,李国正.滚动轴承故障诊断中数据不均衡问题的研究[J].计算机工程与科学,2010,32 (5):150-153Liu T Y,Li G Z.The imbalanced data problem in thefault diagnosis of rolling bearing [J].ComputerEngineering & Science,2010,32 (5): 150-153 (in Chinese) [6] Dai W Y,Yang Q,Xue G R,et al.Boosting for transferlearning [C]//Proceeding of the Twenty-FourthInternational Conference on Machine Learning,2007:193-200 [7] Weiss G M.Mining with rarity: a unifying framework[J].ACM SIGKDD Explorations Newsletter,2004,6(1): 7-19 [8] 蒋少华,桂卫华,阳春华,等.基于核主元分析与支持向量机的监控诊断方法及其应用[J].中南大学学报,2009,40(5): 1323-1328Jiang S H,Gui W H,Yang C H,et al.Method basedon kernel principal component analysis and supportvector machine and its application [J].Journal ofCentral South,2009,40(5): 1323-1328 (in Chinese) [9] 齐咏生,王普,高学金,等.改进MKPCA 方法及其在发酵过程监控中的应用[J].仪器仪表学报,2009,30(12): 2530-2538Qi Y S,Wang P,Gao X J,et al.Application of animproved multi-way kernel principal component analysismethod in fermentation process monitoring[J].ChineseJournal of Scientific Instrument,2009,30 (12): 2530-2538 (in Chinese) [10] Singhal A,Seborg D E.Pattern matching in multivariatetime series databases using a moving window approach[J].Ind.Eng.Chem.Res.,2002,41: 3822-3838
点击查看大图
计量
- 文章访问数: 168
- HTML全文浏览量: 30
- PDF下载量: 4
- 被引次数: 0