owadays, electro hydraulic servo systems are utilized as one of the most popular systems in high-technology industries, such as aerospace, and underwater applications. Because of their application sensitivity, significant attentions have been attracted toward researches about the fault detection and diagnosis of these systems. Obviously, fault detection and diagnosis results in preventing costly failures, increasing safety and reliability, decreasing the shutdown times of the system, avoiding redundant component replacements, and in general raising the performance of the system. The case study of this research is an electro hydraulic servo system, containing a double acting cylinder, a double stage flapper-nozzle servo valve, and the associated controller. The objective of this research is to provide a software-redundant model-based fault detection and diagnosis system for two main common failures of this system, including dropping the bulk modulus due to aeration, and internal leakage. The nonlinear state space equations of the system are linearized using Taylor series. Because of the nonlinearity of the state space equations, the extended Kalman filter (EKF) algorithm is employed to estimate the system states. Condition monitoring of the system is performed in two steps, including fault detection and fault diagnosis. In the first step, the residual between the measured and expected healthy responses of the system is computed by summing the absolute errors between estimated variables and measured values. It is expected that the residual behavior deviates from the natural behavior of the s ystem in the presence of a fault. Residual evaluation and fault detection is accomplished by means of Wald’s sequential test. When bulk modulus dropping and/or internal leakage happen in the system, it is observed that the accumulated values of the residual exceed the threshold value of Wald’s sequential test, while these values remain under threshold in natural healthy operation. In the second step, bulk modulus dropping is diagnosed by online estimation of bulk modulus. To achieve this objective, the bulk modulus estimation is accomplished by adding it as a redundant variable to the state space variables, and observing all the states by means of another EKF. By considering the piston position as the measured output of the system, it is revealed that the bulk modulus is not correctly estimated, because the rank of the observability matrix is not equal to the number of state space variables; hence the system is not observable. To overcome this problem, the piston velocity and the pressures at two chambers of the cylinder are also measured as outputs. In this case, the system is observable because the observability matrix is full rank, and the bulk modulus is correctly estimated. The estimated values of the bulk modulus successfully converged to the actual one in healthy operation of the system, as well as in the faulty scenario of bulk modulus dropping. However, it is observed that the regular EKF is not able to correctly estimate bulk modulus in the presence of internal leakage. It means that regular EKF is not able to correctly distinguish between bulk modulus dropping failure and internal leakage failure, which is not acceptable for the fault diagnosis system. To overcome this raised problem, a promoted algorithm of extended Kalman filter with unknown input is designed and utilized. Using this technique, the bulk modulus is correctly estimated in healthy state, in bulk modulus dropping failure, in internal leakage failure, and in the combination of both faults. Keyword: Extended Kalman filter; Fault detection; Fault diagnosis; Servo valve