This paper presents the theoretical basis of a novel redundancy management procedure developed for fault detection and isolation (FDI) in strategic processes such as spacecraft, aircraft, and nuclear plants where multiply-redundant measurements are available for individual variables. The set of redundant measurements may comprise both direct sensor outputs and analytically derived measurements. The redundancy management procedure presented in this paper is essentially independent of the fault detection strategy and measurement noise statistics, and builds upon the concept of partitioning the set of measurements into “consistent” and “inconsistent” subsets for purposes of estimation and fault isolation, respectively. The proposed procedure is suitable for real-time applications using commercially available microcomputers and its efficacy has been verified on-line in operating nuclear reactors.