航空发动机故障样本有限,利用传统的统计识别方法故障诊断,正确率不高。支撑向量机能解决小样本的故障分类识别问题。研究Support Vector Machine(简称SVM)核函数对识别精度的影响,并把SVM与最大似然法、马氏距离法、最小距离法进行比较,结果表明SVM核函数对故障识别正确率影响不大,基于SVM的航空发动机故障诊断精度高于传统的统计识别方法。 Aero-engine fault samples are limited, the use of traditional statistical methods to identify fault diagnosis, the correct rate is not high. Support vector machine can solve the problem of small sample classification. The influence of Support Vector Machine (SVM) kernel function on recognition accuracy is studied. Compared with maximum likelihood method, Mahalanobis distance method and minimum distance method, the results show that SVM kernel function has little effect on fault recognition accuracy, The precision of aero-engine fault diagnosis based on SVM is higher than that of traditional statistical identification methods.