This paper presents robust sensor fault detection in a flight control system for a satellite Launcher using longitudinal mode and fault tolerant control under such failure conditions. This robust fault detector uses Kalman estimators, Fault Detector and Threshold generator. Kalman estimators estimate the system states. The residuals and their derivatives are used as decision functions to detect the fault. A new method with neural network based threshold generator is proposed to give the threshold values for the decision functions based on the modeling errors in system parameter. This increases the robustness of the fault detection with respect to system parameters uncertainties. If any failure is identified, the control law is modified accordingly using the estimated value as the substitution for the failed sensor output to implement the fault tolerant control. The failure of one of the sensor is considered in this paper. The Results show that the system is able to detect the fault perfectly even with modeling errors introduced in some system parameters. © 2017 IEEE.