Aimed at the problem of the previous fault-tolerant system and that the ordinary least squares support vector machine (LS-SVM) can not effectively perform local detection when a trouble appears in the inertial navigation system of aircraft, a fault detection method based on local weighted LS-SVM is proposed.Firstly, the LS-SVM is locally weighted. Then the innovation of filter by the forecasted innovation of the local weighted LS-SVM regression is used for the test statistics. At last,based on the ADS/GNSS system platform, a simulation is done and analyzed. The result shows that the local weighted LS-SVM fault detection algorithm is good in detection effect, and by using this algorithm the fault detection time and fault isolation rate are reduced greatly.