A scaling factor identification based self-tuned feedback neural network fault-tolerant algorithm is proposed for space robot systems with partial loss of actuator effectiveness. Firstly, a conventional neural network control algorithm is designed for the fault-free system. Then, the real scaling factors are identified by using the scaling factor observer. Finally, a self-tuned feedback neural network fault-tolerant control algorithm is obtained by combining the above neural network control algorithm with the identified scaling factors. The stability criteria of observers and controllers are given strictly based on Lyapunov function method. Numerical simulation verifies the feasibility of the control method.