Study on fast algorithm of neutron radiation field under complex terrain scenario based on ensemble learning approach.

In this study a prediction algorithm has been proposed to rapidly figure out neutron radiation field for nuclear explosion under complex terrain scenario based on ensemble learning approach, which could be an impossibility for traditional radiation transport simulation methodology. By analyzing the influence of complex surface morphology on the radiation field, a series of characteristic parameters which could characterize the topographic features and their influence on the transport of neutrons and secondary gamma in the atmosphere have been extracted with the application of DEM, and the sample sethas been constructedwith the MC simulation results of terrain samples generated by random algorithm, to be used to train the prediction model for the neutron radiation field of nuclear explosion. In order to verify the actual prediction performance of the model, the study has implemented the prediction for the neutron flux, neutron tissue dose and secondary gamma tissue dose under the authentic urban and mountainous terrain scenarios, and analyzed and compared the results from fast prediction and MC simulation in different evaluation dimensions. The comparisons suggest that both of the results are in good agreement with each other, demonstrating that the fast prediction models preliminarily possess the engineering application value. In addition, a feasible approach to improve the generalization performance of the prediction model for various radiation scenarios has been proposed, which could be deemed as a reference for further research.