Reliability based criteria are quite popular for optimal sensor network design. We present a modified definition of system reliability for sensor network design for two applications: reliable estimation of variables in a steady state linear flow process, and reliable fault detection and diagnosis for any process. Unlike the weakest-link based definition of system reliability in the literature, the proposed definition considers the entire system and is consistent with the reliability concept used in classical reliability literature. For each application, dual approaches for defining system reliability are proposed, and their analogy with the reliability problem in the classical reliability literature is established. Using examples and stochastic simulations, the advantage of using the proposed system reliability in contrast to the existing definition is illustrated. Part II of this series of articles presents methods for efficient generation of the system reliability function and its use in optimization-based approaches for designing optimal sensor networks. (C) 2020 Elsevier Ltd. All rights reserved.