This paper presents a novel fault detection method based on incremental locally linear embedding (I-LLE) to improve the accuracy of fault detection for a satellite with high-dimensional telemetry data. Firstly, the I-LLE algorithm is introduced, followed by an application on Satellite “TX-1” telemetry to extract the low-dimensional features, which can be used to perform fault detection with statistical indexes. A rapid semi-physical platform for satellite attitude control systems based on PC104 and AD7011-EVA is designed to perform fault simulation, because limited telemetry contains fewer fault patterns, which renders fault simulation for in-orbit satellites difficult. The I-LLE-based fault detection scheme is then employed to detect anomalies in simulation data. Simulation results presented in this paper validate the fault detection scheme.