When mechanical fault is diagnosed by support vector machine (SVM), the classification ef?fect is closely related to the kernel function. As selecting of the kernel function always lacks theoretical guidance, and approximate computation is adopted in learning course, it led to that the classification result is far from being the expected level. The eigenvector was constructed by mutual information and even amplitude indexes; a series of basic SVM was got by AdaBoost; the ensemble SVM was got by screening the basic SVM with rule of diversity, and weighting the basic SVM which satisfied the di?versity requirement. After the ensemble SVM is put into fault diagnosis of diesel engine of tank, the e- valuation of performance and classification results denote that the ensemble SVM has better classifica?tion performance and a higher classification success rate than single SVM.
Key words
information processing /
adaptive /
support vector machine /
fault diagnosis
{{custom_keyword}} /
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
References
[1] 边肇祺,张学エ.模式识别[M].北京：清华大学出版社,2000: 296-304.
BIAN Zhao-qi, ZHANG Xue-gong. Pattern recognition[M]. Bei?jing： Tsinghua University Press, 2000:296 — 304. (in Chinese)
[2] 胡桥，何正嘉，张周锁，等.基于小波包变换和集成支持失量机 的早期故障智能诊断[J].机械工程学报,2006,42(8) :16-22. HU Qiao, HE Zheng-jia, Zhang Zhou-suo, et al. Intelligent diagno?sis for incipient fault based on lifting wavelet package transforma?tion and support vector machine ensemble[j]. Journal of Mechan?ical Engineering, 2006,42(8) ： 16 - 22. (in Chinese)
[3] Robert E S. A brief introduction to boosting[C]. Proceedings of the Sisteenth International Joint Conference on Artifical Intelli?gence, 1999:1 一 6.
[4] Yoav F, Robert E S. A decision theoretic generalization of on-line learning and an application to boosting[J]. Journal of Computer and System Sciences, 1997，55(1) ： 119 - 139.
[5] 吴涛.核函数的性质、方法及其在障碍检测中的应用[D].长 沙:国防科技大学，2003.
WU Tao. Kernels’ properties, tricks and its applications on ob?stacle detection [ D]. Changsha： National University of Defense Technology, 2003. (in Chinese)
[6] Ben-Hur A, Horn D, Siegelmann H T, et al. Support vector clus- tering[j]. Journal of Maching Learning Research, 2001 (z) ： 125 -137.
[7] Prem M, Raymond J M, Creating diversity in ensembles using ar?tificial data[j]. Information Fusion, 2005,16(1) ： 99 — 111.
[8] John P. Probabilistic outputs for support vector machines and com?parisons to regularized likelihood methods [ EB/ OL ]. Mas?sachusetts： MIT Press, 1999. http： //research, microsoft.com/ —jplatt.
{{custom_fnGroup.title_en}}
Footnotes
{{custom_fn.content}}
