Abstract: Aiming at the problem that the traditional parity space method cannot identify fault component in a gyro-quadruplet of the inertial navigation system, a two-level fault diagnosis method assisted by a support vector machine is proposed to ensure fault tolerance of the gyro module. The generalized likelihood ratio method, whose chi-square volume is constructed by parity residuals, is used for real-time fault detection of the four-axis gyro module. Then, a fault classifier composed of a wavelet packet transform and a support vector machine is constructed to analyze the energy characteristics of gyro output and determinate the operating state of gyro component. Finally, combining the detection results of the generalized likelihood ratio method with the output of the fault classifier, the fault source in the gyro module is detected and isolated. The simulation results show that the proposed method can quickly and accurately identify the faulty device, and thus ensure the performance of the gyro module under a minimum redundancy configuration.

Key words: support vector machine; gyro-quadruplet; integrated navigation system; fault diagnosis