This study addresses the event-triggered fault detection for a class of discrete-time systems with unknown but bounded external disturbance and measurement noise. In order to save limited network bandwidth, an adaptively adjusted event-triggered scheme is designed by considering the probability of system failure and the fault detection progress. Based on the event-triggered system output, an H _ / L ∞ fault detection observer is well designed in finite frequency ranges, such that the generated residual is sensitive to injected faults and robust against disturbance and noise. Moreover, a novel zonotope-based residual evaluation strategy is constructed to reduce conservatism by considering the impacts of external disturbance, measurement noise and measured error caused by event-triggered scheme. Finally, simulation results are provided to demonstrate the effective performance of the proposed event-triggered fault detection strategy.