Load following of the Modular High-Temperature Gas-Cooled Reactor (MHTGR) under Control Rod Drive Mechanism (CRDM) faults and disturbances remains a major challenge. This paper focuses on proposing a fixed-time fault-tolerant control method for this issue without considering the sensitivities associated with parameter setting. Firstly, to reconstruct some unmeasurable states of the MHTGR and the values of CRDM faults, an adaptive learning observer is established. Based on the learning characteristic of Radial Basis Function Neural Networks (RBFNN), the lumped uncertainties can be approximated....