The Age Identification Of Plutonium By RBF Network

Plutonium age is one of the important parameters to infer the history and source of plutonium.The technology of plutonium dating is widely used in the field of nuclear non-proliferation,counter-terrorism and nuclear disarmament.Therefore,it is significant to study the identification method of plutonium age.The purpose of this paper is that using the method of artificial neural network to realize the quantitative identification of the age of plutonium nuclear material by?spectrum fingerprint.Radio-nuclides can produce?-rays in the decay process,the distribution of?-rays intensities with the?-ray energy is called?spectrum.Each radioactive nuclide or nuclear material has a unique?spectrum,so?spectrum can be defined as“?spectrum fingerprint”.The artificial neural network is a kind of intelligent analysis and pattern recognition technology by imitating human brain,which has a good ability of inductive reasoning,nonlinear mapping and distributed parallel data processing.RBF neural network is trained by local approximation,which has the characteristics such as fast convergence speed,high precision and small network scale.In this paper,the nonlinear mapping between the?spectrum fingerprint and the age of the plutonium nuclear material is established by using RBF neural network,and the quantitative identification of the age of plutonium nuclear materials based on?spectrum fingerprint is realized.Plutonium nuclear material contains a variety of nuclides:²³⁸Pu,²³⁹Pu,²⁴⁰Pu,²⁴¹Pu,²⁴²pu,²⁴¹Am and ²³⁷U,in which ²⁴¹Am and ²³⁷U are the products from ²⁴¹Pu.In this work,the plutonium material is chosen which have the same composition and different content as the research object.Firstly,?spectrum of the plutonium nuclear material is simulated by Monte-Carlo method as the radiation fingerprints with different ages.Secondly,?spectrum fingerprint of plutonium nuclear material is taken as the input quantity;the age of the plutonium nuclear material and the nuclide content are taken as output;the artificial neural network is trained through the training sample set.Finally,the trained neural network is proved through the validation sample set.The main research contents are as follows:?1?The age identification of plutonium materials with single nuclide content;?2?The age identification of a group of plutonium materials with the same nuclidecomposition and the different nuclide content;?3?Influence of the content changes with nuclide composition ²³⁹Pu and ²⁴¹Pu.The results show that for the single plutonium nuclear material and multiple groups of plutonium nuclear materials of the same nuclide composition and different content,the age recognition errors are less than 10%;for the three groups of plutonium nuclear materials,when the deviation of the content of ²³⁹Pu is less than 0.6%,4%or 6%,the age recognition errors are less than 10%;when the deviation of the content of ²⁴¹Pu is less than 0.01%,0.01%or 0.06%the age recognition errors are less than 10%.The study shows that the age recognition results are credible for the plutonium materials with the same nuclide content as the training samples,and for the unknown plutonium nuclear materials of different nuclide contents as training samples,the reliability of age recognition results can be judged by the results of the contents of the nuclides.When the nuclide content is within a certain range,the age recognition results can reach the expected accuracy and are credible.