Parity Space-Based Sensor Fault Detection for Traction Rectifier in High-Speed Train

The issue of parity space-based sensor fault detection is solved for traction rectifier of high-speed train, effectively decoupling the effects of unknown initial states and control inputs. Unlike most existing results, process noise and measurement noise are classified as unknown disturbances, which are considered in the mixed logical dynamic model of the rectifier. Subsequently, a linear discrete time-varying dynamic model of rectifier is acquired including unknown disturbances, which is used to implement the fault detection algorithm. The evaluation function calculation of parity space-based fault detection is transformed into solving the minimization of a positive definite quadratic form. By designing observer to construct a new parity relationship, the recursive operation of the residual evaluation function is implemented to reduce the online calculation cost. Finally, an innovation reorganization and singular value decomposition aided parity space algorithm is given, the effectiveness of the proposed method is verified by the simulation experiments.