An Application of Linear Superposition to Estimating Lattice-Physics Parameters

A linear superposition model (LSM) for the speedy and accurate estimation of lattice-physics parameters during within-bundle "pin-by-pin" loading optimization calculations of light water reactor nuclear fuel assemblies has been developed. The LSM has been implemented into the FORMOSA-L code, and typical results show that the run-time requirements can be reduced by at least an order of magnitude relative to performing direct lattice-physics evaluations with the CPM-2 or CASMO-3 code. Moreover, the speedups noted include all overhead expenses associated with the direct lattice-physics calculations required to construct the LSM sensitivity libraries. Additionally, accuracy improvements to the LSM are achieved by inclusion of higher-order cross terms and via quadratic interpolation when perturbing continuous variables. Also, it is shown that the errors generated by this first-order accurate technique can be kept well under control by treating material and spatial shuffles separately during optimizations. The results obtained indicate that the LSM can effectively substitute for direct lattice-physics evaluations throughout the entire optimization process without noticeable loss of fidelity. Finally, both synchronous and asynchronous implementations of parallel computing via the remote-procedure-call approach have been studied to further speed up the creation of LSM sensitivity libraries within FORMOSA-L.