振动与冲击
振動與遲擊
진동여충격
JOURNAL OF VIBRATION AND SHOCK
2015年
11期
133-138
,共6页
沈虹%赵红东%梅检民%曾锐利
瀋虹%趙紅東%梅檢民%曾銳利
침홍%조홍동%매검민%증예리
三阶累积量%图像纹理特征%灰度共生矩阵%特征提取
三階纍積量%圖像紋理特徵%灰度共生矩陣%特徵提取
삼계루적량%도상문리특정%회도공생구진%특정제취
three-order cumulant%image texture feature%gray level co-occurrence matrix%feature extraction
针对发动机不同部位的机械故障特征容易混淆,且往往淹没在其他分量和强噪声中难于区分和提取的问题,提出了一种基于高阶累积量图像特征的柴油机故障诊断方法。截取柴油机6个工作循环的振动信号分别进行三阶累积量计算,累加平均得到1个工作循环信号的三阶累积量,提取柴油机不同故障状态基于三阶累积量图像灰度共生矩阵的图像纹理特征参数,利用支持向量机进行模式识别。试验结果表明:该方法既能抑制噪声干扰,又能充分利用高阶累积量图像中的纹理特征信息分析非稳态信号,提取的特征参数能有效识别发动机6种技术状态,与传统的基于高阶累积量的特征提取相比,提高了故障诊断准确率。
針對髮動機不同部位的機械故障特徵容易混淆,且往往淹沒在其他分量和彊譟聲中難于區分和提取的問題,提齣瞭一種基于高階纍積量圖像特徵的柴油機故障診斷方法。截取柴油機6箇工作循環的振動信號分彆進行三階纍積量計算,纍加平均得到1箇工作循環信號的三階纍積量,提取柴油機不同故障狀態基于三階纍積量圖像灰度共生矩陣的圖像紋理特徵參數,利用支持嚮量機進行模式識彆。試驗結果錶明:該方法既能抑製譟聲榦擾,又能充分利用高階纍積量圖像中的紋理特徵信息分析非穩態信號,提取的特徵參數能有效識彆髮動機6種技術狀態,與傳統的基于高階纍積量的特徵提取相比,提高瞭故障診斷準確率。
침대발동궤불동부위적궤계고장특정용역혼효,차왕왕엄몰재기타분량화강조성중난우구분화제취적문제,제출료일충기우고계루적량도상특정적시유궤고장진단방법。절취시유궤6개공작순배적진동신호분별진행삼계루적량계산,루가평균득도1개공작순배신호적삼계루적량,제취시유궤불동고장상태기우삼계루적량도상회도공생구진적도상문리특정삼수,이용지지향량궤진행모식식별。시험결과표명:해방법기능억제조성간우,우능충분이용고계루적량도상중적문리특정신식분석비은태신호,제취적특정삼수능유효식별발동궤6충기술상태,여전통적기우고계루적량적특정제취상비,제고료고장진단준학솔。
Different positions'mechanical fault features of diesel engines are easy to be confused and they are often drowned in other components and color noises,so it is difficult to distinguish and extract them.Here,a fault diagnosis method based on high-order cumulant image features was proposed.Three-order cumulants for six cycles of vibration signals were calculated,respectively and the results were averaged to get three-order cumulant of one cycle.The image texture feature parameters based on three-order cumulant image gray level co-occurrence matrices (GLCM).For different fault states of diesel engines were extracted.The pattern recognition was performed with a support vector machine (SVM).The results showed that this method can inhibit noises and make full use of texture feature information of high-order cumulant image to analyze unsteady signals,the extracted features can be used to distinguish 6 technical states of diesel engines effectively,the fault diagnosis accuracy is improved compared with the traditional feature extraction based on high-order cumulant.