摘要
为解决预置谱未知的解谱问题,本文首次提出一种广义神经网络(GRNN)算法和迭代算法结合进行的两步解谱法,自主开发了GRNN解谱和迭代法解谱程序,并对2套程序进行分别验证和整体验证。首先用中国实验快堆(CEFR)的活化法实验数据进行分别验证,结果表明:GRNN的解谱结果与理论谱相比,在中子能量大于0.1 MeV时,最大偏差为10.36%,迭代法的解谱结果与最小二乘法的解谱结果最大偏差为9.15%,计算的单核反应率与实验值最大相对偏差为11.71%,符合较好;且与无准确预置谱的迭代法解谱结果相比,GRNN解谱精度更高。最后用俄罗斯碳化硼辐照数据进行整体验证,结果表明:在快中子区域,两步解谱法的结果与有预置谱的迭代法解谱结果最大偏差为11.42%。因此,采用两步解谱法解决预置谱未知的解谱问题是可行的,误差也在可以接受的范围内。本文提出的新型解谱法可为新型堆的解谱提供新的思路,并针对未知预置谱的解谱试验具有一定的参考价值。
To address the challenge of unknown preset spectra,this paper introduces a two-step spectrum unfolding method that combines the generalized regression neural network(GRNN)and the iterative algorithm.We have independently developed the spectrum unfolding codes for GRNN and iteration and conducted separate and comprehensive validations of the codes.Initially,we utilized activation method data from the Chinese Experimental Fast Reactor(CEFR)to validate the codes.The results indicated that at neutron energies greater than 0.1 MeV,the GRNN results deviated by a maximum of 10.36%from the theoretical spectra.The iterative method’s results deviated by a maximum of 9.15%compared to those obtained using the least squares method.The calculated single nuclear reaction rates showed a maximum relative deviation of 11.71%from the experimental values,indicating good agreement.Furthermore,the GRNN method demonstrated higher accuracy compared to the iterative method without accurate pre-set spectra.Finally,comprehensive validation was performed using Russian boron carbide irradiation data,revealing a maximum deviation of 11.42%in the fast neutron region between the two-step method and the iterative method with pre-set spectra.Therefore,employing a"two-step spectrum unfolding method"to address the challenge of unknown pre-set spectra is feasible,with errors remaining within acceptable limits.The innovative spectrum unfolding method introduced in this paper offers fresh perspectives for the spectrum unfolding of new reactors and offers significant reference value for experiments with unknown pre-set spectra.
作者
胡晓
黄毅
王杰
Hu Xiao;Huang Yi;Wang Jie(China Institute of Atomic Energy,Beijing,102413,China;Wuhan Second Ship Design and Research Institute,Wuhan,430064,China)
出处
《核动力工程》
EI
CAS
CSCD
北大核心
2024年第6期1-8,共8页
Nuclear Power Engineering
关键词
两步解谱法
广义神经网络(GRNN)
迭代法
预置谱
Two-step spectrum unfolding method
Generalized neural network(GRNN)
Iterative method
Preset spectra
