Background　Since　there　is　a　dynamic　coupling　effect　between　the　space　manipulator　and　the　base,　the　motion　of　the　space　manipulator　will　inevitably　lead　to　the　elastic　vibration　of　the　base.　Due　to　the　thin　air　and　extreme　environment　in　the　space,　the　elastic　vibration　of　the　base　decays　slowly　and　the　actuator　is　prone　to　the　failure　of　efficiency　loss,　which　will　severely　decrease　the　tracking　accuracy　and　stability　of　the　space　robot.　In　addition,　considering　the　design　tolerances,　installation　deviations　and　frictions,　the　inertial　parameters　of　the　system　are　uncertain,　which　will　also　affect　the　control　effect　of　the　actuator.　This　research　is　dedicated　to　achieving　high-accuracy　and　high-stability　control　of　the　elastic　space　robot　with　actuator　faults　and　uncertain　dynamics.　Methods　According　to　the　linear　momentum　conservation　law,　the　Langrangian　dynamics　equation　of　the　system　is　established.　Based　on　the　singular　perturbation　theory,　the　dynamic　model　is　decomposed　into　a　slow-varying　subsystem　denoting　the　trajectory　tracking　motions　of　the　base　attitude　and　the　joints　and　a　fast-varying　one　denoting　the　nonlinear　vibration　of　the　elastic　base.　A　hybrid　controller　consisted　of　a　decentralized　neural　network　fault-tolerant　controller　of　the　slow-varying　subsystem　and　a　proportion　differentiation　(PD)　feedback　controller　of　the　fast-varying　subsystem　is　formulated.　Conclusions　The　neural　network　fault-tolerant　controller　of　the　slow-varying　subsystem　can　compensate　the　unknown　actuator　faults　and　uncertain　dynamics　and　obtain　H　infinity　convergence　performance,　while　the　PD　feedback　controller　of　the　fast-varying　subsystem　can　damp　out　the　residual　vibration　of　the　elastic　base.　Simulation　results　confirm　the　feasibility　and　effectiveness　of　the　hybrid　control　strategy.