To improve the reliability of fin stabilizer control system, a fault diagnosis algorithm for the fin stabilizer control system based on UIO is presented. The algorithm improves the deficiency compared with the full order Luenberger fault diagnosis observer when the unknown disturbance occurs in the system. Using the optimal decoupling principle, the UIO fault diagnosis observer which can be decoupled to the unknown disturbances is designed. The existence of the UIO necessary and sufficient condition is deduced and proved for the fin stabilizer control system, and then the UIO fault diagnosis observer design steps are listed. The validity of the algorithm based on the NJ5 fin stabilizer control system is verified. The simulations results show that the proposed UIO optimal fault diagnosis observer has the less state estimation error and faster convergence compared to traditional methods, and it can improve the effect of residual convergence and make a quick response to the system failure in the stage of fault diagnosis.