基于支持向量机的航空发动机转静碰摩部位诊断规则提取
Aero-engine rotor-stator rubbing positions diagnosis rule acquisition based on support vector machine
-
摘要: 引入基于支持向量机(SVM)的数据挖掘技术,提出了基于SVM的转静碰摩部位诊断知识获取.首先,基于带机匣的航空发动机转子实验器,模拟了4个碰摩部位的碰摩实验,利用机匣4个部位的应变测试,获取了4个碰摩部位和4个测点的大量实验数据;然后提出了一种基于支持向量聚类(SVC)的诊断知识规则提取方法.在该方法中,利用SVC算法得到特征选取后样本的聚类分配矩阵,最后根据聚类分配矩阵构建超矩形规则.为使规则更加简洁,易于解释,采用规则合并、维数约简、区间延伸等方法对超矩形规则进行进一步简化.利用基于SVM的数据挖掘方法,从大量的碰摩部位实验数据中提取出了转静碰摩部位诊断的知识规则,并进行了相应解释和验证,规则识别率达到了99%以上,表明了该方法的正确有效性.Abstract: The data mining technology based on support vector machine(SVM) was introduced,and aero-engine rotor-stator rubbing positions diagnosis rule acquisition was proposed based on SVM.Firstly,the rubbing experiment of 4 rubbing parts was simulated based on aero-engine rotor tester with the casing,and a large number of experimental data were obtained by using the strain test of 4 parts of the casing.A new approach was proposed to extract knowledge rules from support vector clustering (SVC).Then SVC algorithm was adopted to get the clustering distribution matrix of the sample data with chosen features.Secondly,hyper-rectangle rules were constructed on the basis of the clustering distribution matrix.In order to make the rules more concise and explainable,hyper-rectangle rules should further simplified by using such means as rules merger,dimension reduction and interval extension.Finally,using the data mining method based on SVM,aero-engine rotor-stator rubbing positions diagnosis rules were extracted from a large number of rubbing position experiment data,then explanation and validated accordingly.The recognition rates were more than 99%,showing that the method is corrective and effective,and embodies great practical values.
-
[1] 王德友.发动机转静碰摩振动特性的提取与理论研究[D].北京:北京航空航天大学,1995. WANG Deyou.Research and extract of vibration features on rubbing between rotor and stator for aero-engine [D].Beijing:Beijing University of Aeronautics and Astronautics,1995.(in Chinese) [2] Wang Q,Chu F.Experimental determination of the rubbing location by means of acoustic emission and wavelet transform[J].Journal of Sound and Vibration,2001,248(1):91-103. [3] Han Q K,Yu T,Li H,et al.Hybrid model based identification of local rubbing fault in rotor systems[J].Key Engineering Materials,2005,293-294:355-364. [4] Bachschmid N,Pennacchi P,Vania A.Identification of multiple faults in rotor systems[J].Journal of Sound and Vibration,2002,254(2):327-366. [5] Chu F,Lu W.Determination of the rubbing location in a multi-disk rotor system by means of dynamic stiffness identification[J].Journal of Sound and Vibration,2001,248(2):235-246. [6] HAN Jiawei,Kamber M.Data mining:concepts and techniques[M].San Francisco,US:Morgan Kaufmann Publishers,2001. [7] Mannila H.Data mining:machine learning,statistics,and databases [C]//Proceedings of Eight International Conference on Scientific and Statistical Database Systems.Los Alamitos,US:IEEE Computer Society Press,1996:2-9. [8] FU Chaoyang,ZHENG Jiashen,ZHAO Jingmao,et al.Application of grey relational analysis for corrosion failure of oil tubes[J].Corrosion Science,2001,43(5):881-889. [9] Fu L.Knowledge discovery based on neural networks[J].Communications of the ACM,1999,42(11):47-50. [10] Ziarko W.Discovery through rough set theory[J].Communications of the ACM,1999,42(11):55-57. [11] Quinlan J R.Induction of decision trees[J].Machine Learning,1986,1(1):81-106. [12] 张英.基于支持向量机的过程工业数据挖掘技术研究[D].杭州:浙江大学,2005. ZHANG Ying.A study on process industrial data mining on support vector machine [D].Hangzhou:Zhejiang University,2005.(in Chinese) [13] Vapnik V.Statistical learning theory[M].New York:John Wiley & Sons Inc,1998. [14] Barakat N,Bradley A P.Rule-extraction from support vector machines:a review[J].Neurocomputing,2010,74(1/2/3):178-190. [15] Zhang Y,Su H,Jia T,et al.Rule extraction from trained support vector machines [C]//Proceedings of Ninth Pacific-Asia Conference,Advances in Knowledge Discovery and Data Mining.Berlin,Germany:Springer Berlin Heidelberg,2005:61-70. [16] Ben-Hui A,Hom D,Sidgelman H T.Support vector clustering[J].Journal of Machine Learning Research,2002,2:125-137.
点击查看大图
计量
- 文章访问数: 6749
- HTML浏览量: 12
- PDF量: 13128
- 被引次数: 0