中国电机工程学报
中國電機工程學報
중국전궤공정학보
ZHONGGUO DIANJI GONGCHENG XUEBAO
2014年
32期
5788-5795
,共8页
付文龙%周建中%李超顺%肖汉%肖剑%朱文龙
付文龍%週建中%李超順%肖漢%肖劍%硃文龍
부문룡%주건중%리초순%초한%초검%주문룡
支持向量数据描述(SVDD)%K近邻(KNN)%模糊阈值%不平衡%故障诊断
支持嚮量數據描述(SVDD)%K近鄰(KNN)%模糊閾值%不平衡%故障診斷
지지향량수거묘술(SVDD)%K근린(KNN)%모호역치%불평형%고장진단
support vector data description(SVDD)%K nearest neighbor(KNN)%fuzzy threshold%imbalance%fault diagnosis
水电机组振动故障诊断中常面临样本稀缺及分布不均匀、不平衡等问题,严重影响诊断结果。针对此类问题提出一种基于模糊K近邻(K nearest neighbor,KNN)支持向量数据描述(support vector data description,SVDD)的故障诊断模型。首先利用核变换将故障样本映射到高维特征空间,并采用 SVDD 提取不平衡故障样本域的边界支持向量样本,构建基于相对距离模糊阈值和KNN的决策规则,最终在此基础上建立机组故障诊断模型。用该模型对经过不平衡处理的国际标准测试数据样本进行测试实验,并与支持向量机(support vector machine,SVM)及目前应用较多的SVDD模型的分类结果进行对比,结果表明该模型可有效解决不平衡样本分类倾斜性问题。最后,将模型用于某水电厂机组振动故障诊断,取得了较高的诊断精度,证明了该方法的有效性。
水電機組振動故障診斷中常麵臨樣本稀缺及分佈不均勻、不平衡等問題,嚴重影響診斷結果。針對此類問題提齣一種基于模糊K近鄰(K nearest neighbor,KNN)支持嚮量數據描述(support vector data description,SVDD)的故障診斷模型。首先利用覈變換將故障樣本映射到高維特徵空間,併採用 SVDD 提取不平衡故障樣本域的邊界支持嚮量樣本,構建基于相對距離模糊閾值和KNN的決策規則,最終在此基礎上建立機組故障診斷模型。用該模型對經過不平衡處理的國際標準測試數據樣本進行測試實驗,併與支持嚮量機(support vector machine,SVM)及目前應用較多的SVDD模型的分類結果進行對比,結果錶明該模型可有效解決不平衡樣本分類傾斜性問題。最後,將模型用于某水電廠機組振動故障診斷,取得瞭較高的診斷精度,證明瞭該方法的有效性。
수전궤조진동고장진단중상면림양본희결급분포불균균、불평형등문제,엄중영향진단결과。침대차류문제제출일충기우모호K근린(K nearest neighbor,KNN)지지향량수거묘술(support vector data description,SVDD)적고장진단모형。수선이용핵변환장고장양본영사도고유특정공간,병채용 SVDD 제취불평형고장양본역적변계지지향량양본,구건기우상대거리모호역치화KNN적결책규칙,최종재차기출상건립궤조고장진단모형。용해모형대경과불평형처리적국제표준측시수거양본진행측시실험,병여지지향량궤(support vector machine,SVM)급목전응용교다적SVDD모형적분류결과진행대비,결과표명해모형가유효해결불평형양본분류경사성문제。최후,장모형용우모수전엄궤조진동고장진단,취득료교고적진단정도,증명료해방법적유효성。
The fault samples of hydro-electric generating unit have always been unevenly or imbalanced distributed, which seriously affects the classification accuracy. In order to overcome this disadvantage, a novel support vector data description (SVDD) algorithm improved with fuzzy K nearest neighbor (KNN) decision was proposed. Firstly, the samples were mapped to high-dimensional feature space with kernel transformation, and SVDD was used to extract support vector samples. Then decision rules based on fuzzy threshold and KNN were determined and the novel SVDD algorithm was realized with the new rules. In order to assess the performance of the novel algorithm, two previous SVDD models based on different decisions and support vector machine (SVM) were applied for the comparison with imbalanced datasets from University of California Irvine (UCI) Machine Learning Repository. The experimental results show that the algorithm proposed can efficiently improve the accuracy of classification for imbalanced and uneven samples. At last, the successful application in the fault diagnosis for hydro-electric generating unit attests the effectiveness of the proposed model.