基于改进典型相关分析的中低速悬浮系统异常检测方法

Anomaly Detection Method of Middle-low Speed Suspension System Based on Improved Canonical Correlation Analysis

摘要:

利用悬浮系统的多类监测数据，提出了一种基于改进典型相关分析（canonical correlation analysis， CCA）的中低速悬浮系统异常检测方法。运营线数据验证了该方法能获得较好的阈值，且与基于K-medoids的方法和基于支持向量数据域描述（support vector data description， SVDD）的方法相比，该方法能获得更高的检测率。

Abstract:

This paper presents an anomaly detection method for middle-low suspension system based on an improved canonical correlation analysis （CCA） by using the multiple types of monitoring data from the suspension system. A better threshold and a higher detection rate can be obtined by the proposed method in comparison with the thresholds by K-medoids -based method and the support vector data description （SVDD）-based method.