The MOving Object Simulation Environment: A generic framework for building transient models

A framework called the MOOSE (MOving Object Simulation Environment) has been developed for modeling moving components in the presence of diffusion phenomena. The framework focuses on general elliptic and parabolic problems which can be represented on a two dimensional patched Cartesian grid. The idea of a problem solving environment is presented and the MOOSE is developed within this conceptual paradigm using several novel implementation techniques. Code generation, symbolic computation, and high performance spectral solvers are joined within a flexible and unified tool that implements a mesh linking algorithm capable of minimizing errors induced by moving components in close proximity to material discontinuities.;A major case study based on the 1994 reactivity insertion incident which occurred at the McMaster Nuclear Reactor is undertaken. The flexibility, precision, and robustness of the MOOSE framework and algorithms are exercised by this study. The results from the original tech report are verified for higher dimensional cases.;The MOOSE uses techniques that are mathematically simpler than previously accepted non-linear nodal methods used in nuclear engineering, but still capable of easily representing moving components. A concise ruleset for linking moving meshes is presented which is demonstrated by the framework. Error reductions of several orders of magnitude are demonstrated by the MOOSE's multi-resolution moving mesh algorithm over more costly brute force strategies.;The MOOSE constructs linear finite difference models based on symbolic mathematical descriptions supplied by the model designer. Solutions are computed by transforming abstract descriptions into matrix notation compatible with a collection of high performance parallel linear and eigenvalue solvers. Design techniques are presented for the implementation of a patched non-conformal mesh that links groups of sub-meshes, which can move relative to one another. The generation of a sequence of matrices winch model dynamic components using moving meshes that conserve flow at their boundaries, and the performance of the framework when applied to a variety of test cases is discussed.