With the development of agricultural machinery automatic navigation technology, the precision and stability of agricultural machinery positioning are becoming more and more demanding, MIMU has a broad application prospect in the fields of intelligent agricultural equipment because of its small size, low cost and high reliability. However, due to its own working principle, structure, manufacturing level and error in the installation of the device, MIMU signal output accuracy is poor, resulting in fast dispersion of navigation information over time. It has become an important way to improve the overall performance and accuracy of MIMU-based agricultural mechanical navigation system by using combined navigation technology. However, the increase in system sensors means that the complexity and failure rate of the entire combined navigation system are also increasing. Since the traditional residual χ2 test method hasn’t fully considered the characteristic of the progressive change of the soft fault, which may cause the failure detection of the soft fault, an improved fault detection method of interest sequence was created. Based on the traditional innovation sequence fault detection method, the statistical magnitude with the fuzzy membership function was detected. By using the weighted average algorithm, it can take the detection statistics sitting in the overlapping area of the fault probability density curve and the fault-free probability density curve into the fault detection and increase information for soft fault estimation, so the soft fault could be detected early without increasing the false alarm. The simulation test result showed that the application of this method in the federated filter integrated navigation system could effectively increase the fault-tolerant ability of the integrated navigation system to soft fault.