CAJ | 학술논문

经验模态分解( Empirical Mode Decomposition ,EMD)方法可使滚动轴承振动信号根据自身尺度自适应的分解为若干个固有模态分量( Intrinsic Mode Function ,IMF),信息熵能反映系统的不确定程度,滚动轴承发生故障会导致信息熵发生变化,结合EMD与信息熵,提出了EMD空间状态特征谱熵和IMF奇异谱熵,将其作为概率神经网络的特征向量,构建多传感器系统的多个概率神经网络的初级诊断网络。由于概率神经网络累加层输出结果为属于每一种模式的概率值,用概率神经网络的累加层输出结果构建D-S证据理论的mass函数,通过D-S证据理论进行决策级融合诊断。将该方法用于滚动轴承故障模式分类,试验结果表明该方法的可行性与有效性。与单一传感器诊断相比,提高了故障诊断精度。
경험모태분해( Empirical Mode Decomposition ,EMD)방법가사곤동축승진동신호근거자신척도자괄응적분해위약간개고유모태분량( Intrinsic Mode Function ,IMF),신식적능반영계통적불학정정도,곤동축승발생고장회도치신식적발생변화,결합EMD여신식적,제출료EMD공간상태특정보적화IMF기이보적,장기작위개솔신경망락적특정향량,구건다전감기계통적다개개솔신경망락적초급진단망락。유우개솔신경망락루가층수출결과위속우매일충모식적개솔치,용개솔신경망락적루가층수출결과구건D-S증거이론적mass함수,통과D-S증거이론진행결책급융합진단。장해방법용우곤동축승고장모식분류,시험결과표명해방법적가행성여유효성。여단일전감기진단상비,제고료고장진단정도。
A rolling bearing vibration signal can be decomposed into a number of intrinsic mode function ( IMF) adaptively according to its own scale with the empirical mode decomposition ( EMD) method. The information entropy can reflect the uncertainty degree of the system. The rolling bearing failure will change the information entropy. EMD space state feature entropy and IMF singular spectrum entropy are proposed by combining EMD and information entropy. In the multi-sensor system, primary diagnosis network of mul-tiple probabilistic neural networks was constructed and the feature vector of probabilistic neural network is EMD space state feature entropy and IMF singular spectrum entropy. Because output of accumulation layer of the probabilistic neural network are each kind of model of probability values, mass functions of D-S evi-dence theory can be built using these probability values. The fusion diagnosis of decision level proceeded by the D-S evidence theory. Using this method for rolling bearing fault pattern classification, the results of ex-periments show the feasibility and effectiveness of the method. Compared with the single sensor method in the diagnosis, it improved the accuracy of fault diagnosis.