NUCLEAR TECHNIQUES, Volume. 48, Issue 6, 060401(2025)

Neutron and gamma discrimination method based on KAN and MNv4

Yulong WANG1,2,3, Xiufeng WENG2,3, Xiao LIU2,3, Xiang CHEN2,3, Liang SHENG2、*, and Tong LIU1、**
Author Affiliations
  • 1School of Nuclear Science and Engineering, Shanghai Jiao Tong University, Shanghai 200240, China
  • 2Northwest Institute of Nuclear Technology, Xi'an 710024, China
  • 3National Key Laboratory of Intense Pulsed Radiation Simulation and Effect, Xi'an 710024, China
  • show less
    Figures & Tables(17)
    On site photo (a) and schematic diagram(b) of experimental setup
    Waveforms of neutron and γ signals (color online)
    Flow chart of discrimination algorithm
    Histogram of PSD at different trigger levels (a) -4 mV, (b) -8 mV, (c) -12 mV
    PSD histogram of -8 mV trigger levels at different energy levels (a) 2 500 keV, (b) 5 000 keV
    Schematic diagram of data processing for wavelet scattering network
    Comparison of signals before and after preprocessing (color online)
    Schematic diagram of data processing flow for CNN
    Comparison of accuracy and loss between MNv4 and KAN algorithms (a) Loss curves, (b) Accuracy curves
    Confusion matrix of test results (a) MNv4: Test set 1, (b) KAN: Test set 1, (c) MNv4: Test set 2, (d) KAN: Test set 2
    Schematic diagram of FOM for n/γ discrimination
    PSD histograms of different discrimination methods (a) Charge comparison method, (b) MNv4, (c) KAN
    Waveform signals at 0.1 noise level (color online)
    Waveform signals at (-6, -4] mV triggring level
    • Table 1. Comparison of different network models

      View table
      View in Article

      Table 1. Comparison of different network models

      模型

      Models

      测试集1准确率

      Accuracy of test set 1 / %

      FOM:测试集1

      FOM: Test set 1

      是否能并行

      Parallelizable or not

      参数量

      Params

      MNv499.74.07是Parallel2 420 323
      KAN99.74.40否Serial67 840
    • Table 2. Accuracy at different noise levels

      View table
      View in Article

      Table 2. Accuracy at different noise levels

      模型

      Models

      噪声水平Noise levels
      10%15%20%25%30%40%50%

      电荷比较法

      Charge comparison method

      99.8%99.6%99.2%98.1%96.3%91.3%85.3%
      PCA+GMM83.2%81.3%82.7%82.6%82.4%81.1%78.5%
      SCN+MNv499.7%99.7%99.3%99.25%99.3%99.3%99.0%
      SCN+KAN99.7%99.7%99.4%99.15%99.0%99.2%98.4%
    • Table 3. Accuracy at different thresholds

      View table
      View in Article

      Table 3. Accuracy at different thresholds

      模型

      Models

      信号阈值Signal thresholds

      (-6,-4] mV

      (信噪比约9.13 dB

      SNR approximately

      9.13 dB)

      (-8,-6] mV

      (信噪比约9.32 dB

      SNR approximately

      9.32 dB)

      (-10,-8] mV

      (信噪比约9.9 dB

      SNR approximately

      9.9 dB)

      (-12,-10] mV

      (信噪比约10.5 dB

      SNR approximately

      10.5 dB)

      小于-12 mV

      (信噪比约10.6 dB

      SNR approximately

      10.6 dB)

      SCN+MNv499.26%99.36%99.51%99.62%99.36%
      SCN+KAN99.58%99.66%99.68%99.87%99.36%
    Tools

    Get Citation

    Copy Citation Text

    Yulong WANG, Xiufeng WENG, Xiao LIU, Xiang CHEN, Liang SHENG, Tong LIU. Neutron and gamma discrimination method based on KAN and MNv4[J]. NUCLEAR TECHNIQUES, 2025, 48(6): 060401

    Download Citation

    EndNote(RIS)BibTexPlain Text
    Save article for my favorites
    Paper Information

    Category: NUCLEAR ELECTRONICS AND INSTRUMENTATION

    Received: Sep. 6, 2024

    Accepted: --

    Published Online: Jul. 25, 2025

    The Author Email: Liang SHENG (盛亮), Tong LIU (刘彤)

    DOI:10.11889/j.0253-3219.2025.hjs.48.240349

    Topics