In this paper, integrated design of residual generation and evaluation is proposed to fault detection (FD) of networked control systems. Both the imperfect network transmissions and the sampling effects are considered under the assumption that no stochastic network model is available. By deriving a linear discrete time-varying system equivalent to the sampled plant, the continuous-time behaviors of disturbances and faults during sampling intervals are captured. Moreover, the effects of unknown but bounded packet delays and dropouts in the controller-to-actuator link are described by uncertain input delays with finite number of possibilities. A parity relation-based FD module with multiple residuals is constructed for the derived system model. For the constructed FD module, the set of undetectable faults is obtained from integrated analysis of residual generation and evaluation and then designed to achieve its worst-case minimal geometrical size while guaranteeing zero false alarms. Simulation results on a networked three-tank system are provided to show the merits of the proposed integrated design approach. then the statistical properties of the unknown plant inputs could be calculated [4]. Nevertheless, stochastic network models are often costly to be estimated in practice, especially when the network environment is complex and time-varying [6]. Thus, it is of great importance to investigate FD techniques for NCSs in the absence of stochastic network models. With this consideration, Wang et al. [7] and Wan and Ye [8] utilized the bounded information about packet delays and dropouts, instead of stochastic network models, to facilitate FD design for NCSs.Besides, an NCS involving both the continuous-time plant and digital communication networks is actually a sampled-data system [9][10][11]. In some literatures on FD for NCSs such as [4,5,12], the continuous-time disturbance and fault signals over each sampling interval (which are referred to as inter-sample behaviors throughout this paper) were assumed to be constant. Such an assumption is unrealistic especially when the sampling interval is not sufficiently small. To solve this problem, Zhang et al. [13] and Izadi et al. [14], respectively, adopted the lifting technique and the operator theory to capture the effects of continuous-time disturbances and faults on discrete-time residuals in FD design.A typical FD module consists of residual generator and residual evaluator. The former one generates a residual signal, which captures the difference between the observed system behaviors and the normal system model. The latter one indicates the occurrence of faults by comparing the evaluation function of the generated residual against the selected threshold [15]. In the aforementioned FD literatures for NCSs, either the residual generators or the evaluators were separately designed, or only the residual generator design was considered. A critical problem of such separate design strategy is that an optimal residual generator cannot automatically result in an optimal FD m...