Research On Fast Intelligent Multi-Objective Optimization Of Reactor Radiation Shielding

Nuclear reactors will produce a large number of radioactive rays during operation.The neutron and gamma-ray production will bring irreversible ionizing radiation damage to personnel and facilities,so the radiation shielding design is the crucial link of reactor design.Except for taking dose limitations to the concerned areas in radiation shielding design,many additional requirements are also be taken into accounts,such as space efficiency,material weight,structure performance,and manufacturing costs,making radiation design optimization a typical multi-or many-objective optimization problem.The efficiency of the shielding performance assessment directly affects the development of the radiation shielding optimization of the reactor because the complex structure will make traditional methods take much time to assess.Hence the reactor shielding design engineering requires rapidly find and validate a series of less volume,lightweight,stabilized,and economical solutions under dose limitations.However,the empirical or the Monte Carlo method based shielding designing ways are too inefficient to no longer satisfied modern engineering needs.This paper detailly analyze and discuss the critical problems in reactor shielding optimization problems.Firstly,propose two source activity reconstruction methods based on the deep learning method to ascertain unknown sources’ activity information before designing.And use a source reconstruction model for validating those two methods.Secondly,propose an improved reactor shielding quick assessment method,then test it through the primary and secondary shielding models.Thirdly,propose a reactor shielding design intelligence optimization method,improve the adaptation crossover and mutation operator,offer two methods to solve the multi-objective optimization problem in reactor shielding design rapidly,and validate and compare it through the primary and secondary shielding optimization models.Then entirely independently develop a modularized reactor shielding designing software.Finally,to quickly validate the reactor shielding design’s radiation field,achieve the rapid reactor 3-dimension radiation field calculation.The verification and analysis results show that the mean absolute relative error of the two source activity reconstruction method is less than 15% and25%,proving they can accurately reconstruct uncertain sources’ activity;the mean absolute relative error of the reactor shielding performance quick assessment method is maximum to 12% in the primary and secondary shielding model,shows it can rapidly assess shielding design scheme;the comparison result shows the reactor shielding design intelligence optimization method’s speed is improved two orders of magnitude to the pure Monte Carlo based optimization method,reactor shielding multi-objective optimization methods’ Pareto optimal results uniformly distributed in objective space,which confirming they can simultaneously optimizing objectives under dose limitations,achieved the goal of rapidly and intelligently optimizing multi-objective problems in the reactor shielding design;the developed reactor shielding design intelligence optimization platform can automatically complete reactor shielding multi-objective designing;the reactor 3-dimension radiation field calculation program’ results are consistent with MCNP’s,proving the program can analyze and validates the reactor radiation field in a short time.This paper form a self-contained reactor shielding optimization framework.The related research and development achievements can efficiently provide technical support for the development of reactor radiation shielding design and can be applied in reactor shielding design engineering.