| Monte Carlo method is an important tool of high fidelity reactor simulation because of the advantage of being able to accurately modeling complicated geometry and physical phenomena of reactor simulation,but its slow convergence limits its usage in engineering applications.High fidelity reactor particle transport simulation usually requires obtaining fine grained,globally converged calculation results and this is a difficult challenge for Monte Carlo codes.Due to the immense size and thick shielding of nuclear reactors,the convergence rate is extremely non-uniform and will cost a prohibitive run time to obtain global converged results.Especially for fission reactors,there also exist other problems like non-uniform source distribution and massive tally numbers.Based on the super Monte Carlo program for nuclear and radiation simulation SuperMC,this work proposes studies of accelerating methods for the above mentioned difficulties.1)For the problem of non-uniform convergence rate in high fidelity whole core simulation,the work developed a phase space particle density uniformity based adaptive global variance reduction method.Starting from the traditional weight window generator,the work introduced a new phase space cell importance definition and developed a new weight window generator that is capable of optimizing global calculation efficiency.Based on this global weight window generator,this work further studies a global-local coupled variance reduction method:use a global weight window as part of calculation input can significantly improve the performance of linear combination weight window generator,which can accelerate the convergence of several local tally targets.The tests with a simple spent fuel storage pool model shows that the developed method can improve the global calculation FOM more than a thousand times.For a chosen local tally,the global-local coupled method improves its calculation FOM for hundreds times comparing with the analog case.2)For the problem of non-uniform convergence rate in fission reactor simulations,this work developed a global weight window-uniform fission site coupled accelerating method:The uniform fission site method biases the fission source sampling to obtain a more uniform source site distribution while the global weight window further bias the particle transport process.On the other hand,the uniform fission site method also biases the born weight of the fission site,making it consistent with the weight window ranges thus avoiding unnecessary splitting or Russian roulette.Tests with the Hoogenboom reactor model shows that the coupled method improves the calculation efficiency about 30 times for thermal group tallies and 20 times for fast group tallies comparing with the analog case.3)For the problem of massive tallies,this work developed a tree-based method to accelerate tally scoring.This method builds a tally assist tree that all the nodes map one-for-on to the geometry cells.All the tally information read from the input file is stored in the tree.By mapping the geometry information of the cell where the particle is currently in to a node in the tree,the tallies need to be scored is directly retrieved from the node.The test with the Hoogenboom reactor model which contains about 6 million tallies shows that the tally time usage is only 4 percent of the total runtime;the tally efficiency is significantly improved.To test the developed accelerating methods for high fidelity reactor simulation with more realistic and complicated reactor models,this work tested the methods with the C-Lite reference model of the ITER Tokamak reactor and the BEAVRS reactor model.For the C-Lite model,the FOM of shielding calculations increases about hundreds of times.For the BEAVRS model,the FOM of flux distribution calculation increases about 10?20 times.These test results shows that the accelerating methods developed in this work can effectively improve the calculation efficiency of high fidelity reactor simulations. |
PDF Full Text Request