基于小波变换的多尺度主元分析在传感器故障诊断中的应用

Application of MSPCA Based on Wavelet Transform to Sensor Fault Diagnosis

讨论了多尺度主元分析方法在传感器故障诊断中的应用问题.为了解决传统的多尺度主元分析方法不能实现对传感器故障的全面检测问题,本文结合小波变换在相关传感器信号的各个尺度上建立主元分析模型,使这种方法能够同时检测到低频故障和高频故障.实际应用中设计了固定窗长的移动窗口,根据最后一个尺度系数计算残差空间的平方预报误差统计量进行故障检测;在检测到传感器故障后,再采用传感器有效度指标这种具有定量辨识标准的参数对故障传感器进行辨识.最后,通过液体火箭发动机试车台液氢供应系统的传感器故障诊断验证了这种方法的实用性和有效性.

Multi-scale principal component analysis （MSPCA） for sensor fault diagnosis is discussed. To resolve the problem that traditional MSPCA can not detect sensor faults roundly, the principal component analysis models are established by integrating with wavelet transform for the each scale coefficient of correlative sensors. These models can detect the failure with low frequency and that with high frequency simultaneously. In application, a moving window with constant length is designed. After wavelet transform of the data from the window, square prediction error in the residual space is calculated with the latest coefficient to detect failure. Meanwhile, the sensor validity index with quantitative standard is calculated to identify the faulty sensor. Finally, the applicability and effectiveness of the proposed method is illustrated by the sensors fault diagnosis of liquid rocket engine ground testing bed hydrogen providing system.