PCA model was set up based on the normal history data of the system to detect the sensor faults by comparing the projection on the residual subspace to which the real measurement vector and normal vector were projected.

• 主成分分析法根据系统正常的历史数据建立数理统计模型，通过传感器实际测量数据与正常数据在故障子空间投影的比较，对传感器的故障进行在线检测。