Various forms of faults have a significant impact on the performance of navigation systems, especially satellite navigation system faults occur most frequently, so accurate fault detection of navigation systems is particularly important. At present, the commonly used satellite navigation system fault detection algorithms include parity vector method, least squares method and other detection methods. Such methods have good detection effects on large faults with large deviations in measurement information, but the system is slowly reduced. The detection effect of the fault is not obvious. However, the detection and correction of small gradual deviations is an important issue that navigation systems must address in high-precision applications. Aiming at the fact that most fault detection methods are insensitive to slow fault detection, and have large delay, and cannot judge various fault categories, a fault detection algorithm based on parity vector sliding window accumulation improvement is proposed. The method comprehensively utilizes the observations of the current and previous epochs, firstly detecting the current epoch measurement information by using the traditional parity vector method to determine whether the system has a hard fault. On this basis, the sliding window is used to calculate the square of the parity of the first N epochs at the current time, construct a new detection statistic, and solve the detection state in real time by constructing a sliding window. The simulation results show that the proposed method can detect and isolate faults in time when the integrated navigation system has hard faults and slow faults, thus improving the fault tolerance of the system.