CAJ | 학술논문

为了克服传统分类器训练时需要大量特征样本的缺点，设计了一种基于多元支持向量机(MSVM)的离心式压缩机转子故障诊断分类器。支持向量机是一种新的机器学习方法，在小样本和高维二元分类方面有非常突出的优点。实验证明，依据此优点设计的MSVM分类器模型不仅可以在小样本情形下对模型进行快速训练，并且可以快速识别多种故障类型，同时识别率也比传统的神经网络方法有较大提高，从而达到提高设备监测和故障诊断效率的目的。
위료극복전통분류기훈련시수요대량특정양본적결점，설계료일충기우다원지지향량궤(MSVM)적리심식압축궤전자고장진단분류기。지지향량궤시일충신적궤기학습방법，재소양본화고유이원분류방면유비상돌출적우점。실험증명，의거차우점설계적MSVM분류기모형불부가이재소양본정형하대모형진행쾌속훈련，병차가이쾌속식별다충고장류형，동시식별솔야비전통적신경망락방법유교대제고，종이체도제고설비감측화고장진단효솔적목적。
For solving the defect of traditional classificatory with many samples, a new classificatory recognizing faults based on Multi-Support Vector Machine (MSVM) is proposed for centrifugal compressors. SVM is a new machine study method which has excellent advantages in small-sample and multi-dimension binary classification. The new MSVM classificatory can be studied in a few samples rapidly to recognize several kinds of new faults. At the same time, the experiments showed that recognizing correct rate increased more greatly compared with traditional BP method.