In this paper, the sensor fault detection (FD) problem is investigated for the dynamic pointthe-bit rotary steerable system (DPRSS), which is known as state-of-the-art oil drilling equipment. Firstly, the DPRSS dynamic is modelled as a linear system with drilling disturbances, which consists of model perturbations, measurement disturbances, and unknown inputs. Secondly, the finite-frequency H − ∕H ∞ unknown input observer is proposed to generate the residual, which is sensitive to sensor faults and robust against drilling disturbances. Subsequently, considering model perturbations and measurement disturbances, the zonotope-based residual evaluation method is presented, and the FD decision-making logic is designed based on the calculated residual bounds. Finally, numerical simulations and prototype experiments are provided to illustrate the effectiveness of the proposed sensor FD mechanism.