| Most models and tools for analyzing reliability have traditionally been based on binary approximations; these models and tools are expanded in this dissertation to allow multistate and continuum analysis. A series of customer-centered reliability measures, valid for binary, multistate, and continuum reliability models, is defined and developed. Techniques are presented for computing or approximating the state of the system as a function of the states of its components, based on subsystem identification and an application of multiquadric interpolation, and an extension of the interpolation method is presented which will preserve any monotonicity in the original customer-supplied data set. Methods for multistate reliability modeling based on continuous-time Markov chains are developed and extended. Procedures for assessing the accuracy of a continuum model discretization are proposed, and reliability measure computations for non-trivial continuum and mixed systems are illustrated. A comprehensive software package which performs non-binary reliability analyses is included and documented. |
