Research On Design Method Of Reactor Shielding Optimization

Reactors produce neutron and gamma rays during operation.Shielding layer is often used around the core to reduce radiation leakage in engineering.The main purpose of reactor shielding optimization design is to reduce the thickness of the shielding layer,thus reducing the size of the reactor and increasing the shielding efficiency of the shielding layer.At present,the research direction of reactor shielding optimization is mainly divided into the research on new composite materials and the research on the reasonable arrangement of shielding layer structure.Reactor shielding scheme optimization is of great significance for reactor miniaturization and lightweight.Based on the National key research and development program"New Marine Nuclear Reactor Technology"(2019YFB1901300),this paper selected a reactor as the research object for shielding calculation and analysis,the results can provide reference for reactor shielding optimization design.In this paper,the discrete ordinate method(S_Nmethod)of the deterministic method and Monte Carlo method(MCNP method)are used to model and analyze the reactor core and the structure outside the reactor.The main research contents and conclusions of this paper are as follows:1.Three-dimensional neutron transport program MCNP5 was used to model reactor,and reactor critical characteristics and material shielding performance were studied in this paper.The material arrangement of multi-layer shielding is considered.2.The model of the reactor was modeled by the two-dimensional deterministic module"T-NEWT"in SCALE software,and the rationality of the model is verified by the critical experiment benchmark reference questions.The calculation results show that"T-NEWT"agrees well with the calculation results of MCNP.3.SCALE software was used to calculate 1001 data sets.Based on Tensor Flow deep learning framework in Python,the neutron beam intensity model with error correction terms is trained with training sets,and the generated prediction model was verified with test sets.Compared with the model without error correction,the prediction accuracy is improved by40%.4.According to the difference of shielding capacity required by different reactor locations,by setting weights for different optimization targets which based on Python language,the performance of multiple schemes within the corresponding thickness is predicted by using the prediction model,and the best material thickness scheme is found.5.MCNP5 program is used to verify the optimal arrangement of shielding schemes.