CAJ | 학술논문

为了实现对故障轴承的特征提取和对故障特征的准确分类，该文提出了应用小波包变换与改进的多元最小二乘支持向量机(LS-SVM)相结合进行滚动轴承故障模式识别的方法。首先，利用小波包对滚动轴承振动信号进行分解和重构，并构造特征向量；然后，针对传统的单个核函数不能兼顾学习能力和泛化能力的缺点，提出了应用混合核函数对多元 LS-SVM进行改进的算法；最后，将特征向量作为输入，分别应用于常用核函数和基于混合核函数的多元 LS-SVM对滚动轴承故障类型进行仿真判别对比实验。结果证明了所设计方法的有效性。
위료실현대고장축승적특정제취화대고장특정적준학분류，해문제출료응용소파포변환여개진적다원최소이승지지향량궤(LS-SVM)상결합진행곤동축승고장모식식별적방법。수선，이용소파포대곤동축승진동신호진행분해화중구，병구조특정향량；연후，침대전통적단개핵함수불능겸고학습능력화범화능력적결점，제출료응용혼합핵함수대다원 LS-SVM진행개진적산법；최후，장특정향량작위수입，분별응용우상용핵함수화기우혼합핵함수적다원 LS-SVM대곤동축승고장류형진행방진판별대비실험。결과증명료소설계방법적유효성。
To solve the feature extracting and feature classifying of fault bearings diagnosis,a method of fault recognition of rolling bearings based on wavelet packet transform and improved multi-classification least square support vector machine (LS-SVM)was presented.Firstly,rolling bearing vibration signal was decomposed and reconstructed with wavelet packet transform to construct characteristic vectors. Then,in response to contradiction in learning and generalization of traditional single kernel function,an algorithm of applying hybrid kernel function to improve LS-SVM was proposed.Finally,simulation test was finished to realize fault patterns recognition by LS-SVM algorithms based on traditional kernel func-tions and hybrid kernel function respectively,when the characteristic vectors were inputs.The experimen-tal results demonstrate the effectiveness of the method proposed.