Automated instrument design and optimisation using a genetic algorithm running on a distributed processing network

摘要

Monte Carlo codes provide a means for accurately simulating nuclear and optical photon transport and are widely used to design and optimise new nuclear instruments. Typically, a number of design variations are simulated and the best variation is selected for construction and experimental testing. The process is limited by the time required to set up and run each design. Separate aspects of complex designs are often optimised sequentially rather than globally. A software package has been developed that allows users to create parameterised families of instrument designs, which are then searched by a genetic algorithm for the globally optimal design. The package has been interfaced to the MCNP, EGSnrc and PHOTON codes for solving neutron, gamma-ray, electron and optical photon design problems.