针对航空发动机振动监控异常样本少的问题,用单类支持向量机建立了一种振动异常检测模型,在仅对正常数据进行训练的基础上便可以进行发动机振动异常检测工作。根据近期数据的重要性要大于早期数据的重要性这一特性,提出加权单类支持向量机算法,为不同架次的样本赋予不同的权系数。实验分析结果表明了检测模型的有效性。 Aiming at the problem of few samples of abnormal vibration monitoring of aeroengine, a single class support vector machine is used to establish a vibration anomaly detection model, and anomaly detection of engine vibration can be carried out only on the basis of normal data training. According to the fact that the recent data is more important than the importance of the earlier data, a weighted single-class support vector machine algorithm is proposed to give different weights to different samples. Experimental results show the effectiveness of the detection model.