Abstarct: At first, the unmanned aircraft and the importance of its sensors is reviewed. Fault concept and its variants are described. Then methods of fault detection are introduced. The considered model is Machan unmanned aircraft linear system which is assumed multiple inputs- multiple outputs. The most methods that used to fault detection and isolation were baxsed on residuals changes. Residuals do not give enough information about the nature of the fault. There are some imperfect studies about fault detection and isolation, in which fault varies by the time. The method proposed in this thesis using model-baxsed approach that detects and reconstructs also the time-varying faults with designing and using the adaptive observer. In the simulation, the adaptive observer is compared with classical method at first and it is proved that this method is very convenient for time-varying fault detection and diagnosis. Also it is observed that the observer diagnosed faults correctly even though in the presence of several uncertain faults. The aim of this thesis is to obtain the construction of the fault that happens in the system output. This method is considered as a fast fault detection methods with high precision and it is one of the the most widely used and reliable methods in this field.