Abstract: Because of the limitation of Support Vectors Machine（SVM）, this paper studies a fault detective method for aero-engine called Relevance Vector Machine（RVM） which is based on sparse Bayesian learning. The Exhaust Gas Temperature（EGT） of aero-engine is a significant parameter in engine monitoring and fault diagnosis, so the RVM is used to predict it. According to the simulation, though there are just a few samples, it is proved that RVM can predict EGT accurately in time. Meanwhile, the faults can be detected by the relative errors between the real and the predictive.

Key words: Relevance Vector Machine（RVM）, aero-engine, fault diagnosis, Exhaust Gas Temperature（EGT）

摘要： 针对支持向量机算法存在的不足，研究了一种基于稀疏贝叶斯框架的机器学习方法——相关向量机在航空发动机故障检测中的应用。排气温度是进行发动机监控与故障诊断的重要依据，应用相关向量机对其进行预测。通过仿真实验，证明了相关向量机方法在样本数据较少的情况下只产生了很少的相关向量，并且能够及时准确地预测出发动机排气温度；同时可以使用真实值与预测值的相对误差作为系统是否发生故障的判断依据。

关键词: 相关向量机, 航空发动机, 故障诊断, 排气温度