Neural Network Radionuclide Identification Algorithm Based on Exponential Smoothing

The development of nuclear energy and the application of nuclear technology is inseparable from the detection of radionuclides. Considering the incomplete utilization of γ-spectral information in traditional spectral analysis methods, the accuracy is not high. In this study, exponential smoothing is performed on the energy spectrum signal of length 1024 to eliminate the influence of the statistical fluctuation of the detector, and then it is used as the input of the neural network to realize the full spectrum analysis. The method avoids the complex feature extraction process and can identify a single radionuclide in real time.