基于在线SVDD的航空发动机异常检测方法

摘要

针对飞参数据样本量大、分布不均衡且随时间不断积累的特点给航空发动机异常检测带来的问题，提出采用在线SVDD进行航空发动机异常检测的方法。首先介绍了在线SVDD的基本原理，然后采用大规模数据对比分析了现有在线SVDD方法的性能，最后采用两组典型发动机典型异常进行实验。结果表明，在线SVDD能够快速准确地识别发动机异常。%To tackle the aeroengine novelty detection problems resulted from the characteristics of flight data including large scale, unbal-anced distribution and increasing with time, a method of aeroengine novelty detection based on online SVDD is proposed. Firstly, the theory of online SVDD is introduced briefly; secondly, the performances of current online SVDD methods are investigated comparably with large scale dataset; lastly, two typical novelties are used to execute aeroengine novelty detection. Experimental results show that online SVDD can recognize the aeroengine novelty rapidly and accurately.