The Research On Irradiation Effects In Reactor Structural Materials Based On Rate Theory And Machine Learning

Reduced activation ferritic-martensitic(RAFM)steel is one of the important candidate structural materials for fusion reactors because of its good thermophysical properties and irradiation resistance,as well as a relatively mature industrialization basis.However,there are still many challenges in the development and application of RAFM steel.Under neutron irradiation,RAFM steel will produce a series of radiation damage behaviors such as irradiation swelling,irradiation hardening/embrittlement,which seriously affects its service safety.In order to meet the service requirements of future commercial reactor structural materials,it is necessary to accurately evaluate their irradiation damage performance under different conditions,and to provide guidance on the optimization of the irradiation resistance performance of RAFM steel based on understanding the mechanisms.It is the most visual and effective method to study the irradiation effect of reactor structural materials through neutron irradiation experiments.However,neutron irradiation experiments are difficult,expensive and time-consuming to carry out.Moreover,there is currently a lack of experimental conditions for fusion reactor neutron irradiation in the world,and alternative research can only be carried out through fission reactors,spallation neutron sources or heavy ion irradiation experiments.The spatial and temporal distributions of defects produced by these irradiation experiments are different from those produced by neutron irradiation in fusion reactors.Therefore,it is important to develop theoretical models for the study of radiation effects,especially the rate theory method has the advantages of fast and general calculation for high dose conditions under working conditions.However,the current rate theory model is difficult to be solved for high-temperature and high-dose neutron irradiation,and the model cannot give accurate prediction results for special stages such as the incubation period of swelling.Moreover,the influence of the transmutation product helium on the radiation effect brought by the neutron irradiation of fusion reactors also needs further research.To solve the above problems,based on the rate theory and machine learning methods,this research has carried out a series of research on the irradiation effects of reactor structural materials.The main research contents and results are as follows:1)A machine learning-rate theory(ML-RT)coupled model of irradiation swelling in materials was established,and the prediction of the swelling behavior of structural materials under high temperature and high dose conditions was realized.Using this model,the main reasons for the difference in irradiation swelling resistance of austenitic steel and ferritic/martensitic(F/M)steel were compared and analyzed.Based on the irradiation dose data set during the incubation period,the machine learning method was used to predict the dose at the starting point of the linear growth of swelling.Comparing the results calculated by five commonly used machine learning algorithms,it was found that the random forest(RFR)method had the highest coefficient of determination(R),is 0.91.Using the improved model of rate theory,the swelling behaviors of austenitic steel and F/M steel under neutron irradiation at different doses and temperatures were predicted and compared with the experimental results.For austenitic stainless steel AISI 316 and F/M steel JLF-1,the peak swelling temperature under neutron irradiation is about 500℃ and 425℃,respectively.The initial dose of linear swelling is 42 dpa and 34 dpa,respectively.The steady-state swelling rate is about 0.85%/dpa and 0.02%/dpa,respectively.When the irradiation dose reached 150 dpa,the volume swelling of AISI 316 is about 80%.These conclusions are consistent with the experiments and verify the validity of the ML-RT coupling model.The model was used to predict cavity swelling for Chinese low activation ferritic/martensitic steel(CLAM)under neutron irradiation,and its swelling was about 2.5%after 100 dpa.Based on the ML-RT model,a comparative analysis of the irradiation swelling of austenitic steel and F/M steel was carried out.The effects of various factors such as initial dose of swelling linear growth,cascade efficiency,sink strength,point defect diffusion coefficient and point defect absorption bias on irradiation swelling were analyzed.The results show that the absorption bias is the main reason for the difference in the radiation swelling resistance of austenitic steel and F/M steel.The ML-RT coupling model provides an effective method for the prediction,evaluation and mechanism analysis of the swelling behavior of structural materials under neutron irradiation.2)The prediction of the irradiation hardening/embrittlement behavior of RAFM steel and other reactor structural materials is realized by combining the classical strengthening theory and the rate theory model.Based on the analysis of classical strengthening theory,it is recognized that the defect microstructures such as dislocation loops,voids and helium bubbles introduced by irradiation are the main reasons for the irradiation hardening/embrittlement in materials.Using an improved rate theory model,the evolution of the defect microstructure under irradiation was simulated.Using machine learning,the irradiation hardening saturation dose of different structural steels was predicted.For RAFM steel,the irradiation hardening peak temperature is about 300℃,and the lower the temperature,the more serious the hardening.At 300℃,the yield strength increment of RAFM steel has an exponential relationship with the irradiation dose,and reaches saturation at around 20 dpa.At saturation,the yield strength increment exceeds 550 MPa,and the total yield strength is close to 1000 MPa.Among all kinds of irradiation-induced defect microstructures,dislocation loops have the highest contribution to the total yield strength increment,about 80%.Based on the irradiation hardening results,the relationship between the microhardness,the ductilebrittle transition temperature(DBTT)was obtained by using the parametric model.The calculated results were in agreement with the experiments.The irradiation hardening of austenitic stainless steel 316 is compared and analyzed.The difference is mainly due to the difference in the size and density of dislocation loops.Therefore,reducing the generation of dislocation loops under irradiation is an effective way to reduce the irradiation hardening/embrittlement.3)To simulate the high-energy neutron irradiation effects in fusion reactors,a machine learning model considering the helium effects are established.The prediction of the irradiation effect of materials under different helium production rates is realized.The effect of transmutation helium on the irradiation effects is analyzed.Using the model considering the helium effect,the irradiation swelling of RAFM steel under different helium production rates was predicted,and the irradiation swelling of RAFM steel doped with Ni and B under fission neutron irradiation was compared and analyzed.The results show that under the condition of 6.5 appm He/dpa,the cavity swelling of RAFM steel at 50 dpa is about 1.2%,which is nearly two times of the void swelling.After 80 dpa,the cavity swelling exceeds 1.6%,indicating that the irradiation swelling of RAFM steel was more severe with higher helium production rates.The irradiation hardening/embrittlement of RAFM steel under high helium production rate in the spallation neutron sources was predicted using a machine learning model considering the helium effects.The yield strength is 150MPa higher than the results without helium and the irradiation hardening saturation dose was also increased.Based on machine learning,the influence of various characteristic variables on the irradiation effect was compared,and the helium production rate was the one of the most important factors.The irradiation swelling behavior of RAFM under fusion reactor conditions(with helium production rate of about 10 appm He/dpa)is predicted using a rate theory model considering the helium effects.When the irradiation dose reaches 100 dpa,the irradiation swelling is about 1.9%.In summary,this paper combines machine learning,rate theory,and classical hardening models to develop a coupled model for irradiation effects under hightemperature and high-dose neutron irradiation conditions.It provides theoretical support and methods for efficiently evaluating the irradiation performance of reactor structural materials and guiding the optimal design of radiation-resistant materials.