In this paper, we propose a robust fault detection and isolation (FDI) strategy based on zonotopic unknown input observers (UIOs) for discrete-time descriptor linear time-varying (LTV) systems subject to uncertainties and additive actuator faults. System uncertainties including state disturbances and measurement noise are unknown but bounded by predefined zonotopes. The uncertain state estimations and constructed residuals for robust FDI are propagated in a sequence of zonotopes. Based on a defined performance criterion, the fault detection (FD) observer gain is designed to be robust against uncertainties and meanwhile sensitive to faults. The explicit computational method for the FD observer gain is derived. In addition to include fault isolation, a bank of zonotopic UIOs are employed. Finally, we apply the proposed method into two case studies to show its effectiveness.