From FUELCON to FUELGEN: Tools for Fuel Reload Pattern Design

Abstract

FUELGEN is an effective tool for refuelling design, i.e., for solving the in-core fuel management problem at nuclear power plants. Devising good fuel-allocations for reloading the core of a given nuclear reactor, for a given operation cycle, is crucial for keeping down operation costs at plants. Fuel comes in different types, and is positioned in a grid representing the core of a reactor. The starting point was Galperin and Nissan’s prototype which eventually led to FUELCON, a rule-based expert system with the same task. FUELGEN, instead, is based on a genetic algorithm for optimization, and is at the current forefront of research in refuelling design, where genetic techniques are now getting increrasing recognition. The end result of over a decade of research within this sequence of projects yielded a set of alternative, partly overlapping architectures. Nodal algorithms to carry out parameter prediction by simulation, heuristic rules in FUELCON’s ruleset and meta-level refinement ergonomic techniques by which the ruleset can be refined during a session with FUELCON, attempts with neural computation on top of the latter, and then, replacing the ruleset altogether by resorting to genetic algorithms, are the sequence of techniques that were in turn applied, in the development of FUELCON and then FUELGEN. This actually reflects the sequence of emergence of expert systems and then neural computation methods, then genetic and hybrid methods, in knowledge engineering in general and in its application to nuclear engineering in particular.