NaI(TI) Gamma-ray Spectrometry Data Analysis And Software Design

The full energy spectrum measured by NaI(TI) gamma-ray spectrometer has a shortcoming of low energy resolution and there are too many peaks overlapped in low energy area, but the information of peak 609keV photopeak of ²¹⁴Bi, 583keV photopeak of ²⁰⁸Ti and 662keV of ¹³⁷Cs (man-mad nuclide) are very important to us. So it demands a further research on the accurate resolution to completely overlapped peaks in low energy area and the accurate resolution method never comes easy to us.A new Radial Basis Function Neural Network (RBFNN) based on Three Peaks Gaussian (TPG) function is constructed to get accurate resolution of overlapped peaks in low energy area. After do some pretreatments of gamma-ray data (in txt form), we can get the nine parameters of TPG which are from the restriction TPG fit of the ROI data in low energy area. Then use MATLAB to plot three separate Gaussian peaks, from these plots, we can check and save the best fit parameters of TPG model.To form the training sample of RBFNN, we use the data of overlapped peaks in Xiangshan to form a matrix of 8x60, use the matrix to train the RBFNN. Two types of RBFNN are presented to the sample and use a test sample to test the RBFNN. Also an error decreasing curve of RBFNN is showed in the end.The best trained net has the ability of reconstructed structure, so we can get a more accurate resolution of overlapped peaks in low energy area. From the results of resolution, the RBFNN based on TPG used can get a accurate resolution, there also a comparing plot of the three separate Gaussian peaks use nine resolution Gaussian parameters. So the conclusion is that the RBFNN based on TPG is very suitable and effectively to the resolution of completely overlapped peak in low energy area.