Fast neutron-gamma discrimination in organic scintillators via convolution neural network

Yoon, Seonkwang; Lee, Chaehun; Won, Byung-Hee; Hong, Seong-Gil; Seo, Hee; Kim, Hodong

doi:10.1007/s40042-022-00398-x

Cited by 10 publications

(2 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The implementation of neutron/gamma classification ML models on commercial FPGAs is established. Previous examples of work done include but are not limited to [21][22][23][24][25][26]. However, the combined advances in both open-source eFPGAs and translation of ML models to hardware now provide the basis for a viable pathway to a real-time, low-power and compact neutron/gamma classification method using ML models directly on a custom ASIC that integrates all required circuitry, including the eFPGA fabric.…”

Section: Jinst 19 P07034mentioning

confidence: 99%

Investigating resource-efficient neutron/gamma classification ML models targeting eFPGAs

Johnson,

Boxer,

Prakash

et al. 2024

J. Inst.

View full text Add to dashboard Cite

There has been considerable interest and resulting progress in implementing machine learning (ML) models in hardware over the last several years from the particle and nuclear physics communities. A big driver has been the release of the Python package, hls4ml, which has enabled porting models specified and trained using Python ML libraries to register transfer level (RTL) code. So far, the primary end targets have been commercial field-programmable gate arrays (FPGAs) or synthesized custom blocks on application specific integrated circuits (ASICs). However, recent developments in open-source embedded FPGA (eFPGA) frameworks now provide an alternate, more flexible pathway for implementing ML models in hardware. These customized eFPGA fabrics can be integrated as part of an overall chip design. In general, the decision between a fully custom, eFPGA, or commercial FPGA ML implementation will depend on the details of the end-use application. In this work, we explored the parameter space for eFPGA implementations of fully-connected neural network (fcNN) and boosted decision tree (BDT) models using the task of neutron/gamma classification with a specific focus on resource efficiency. We used data collected using an AmBe sealed source incident on Stilbene, which was optically coupled to an OnSemi J-series silicon photomultiplier (SiPM) to generate training and test data for this study. We investigated relevant input features and the effects of bit-resolution and sampling rate as well as trade-offs in hyperparameters for both ML architectures while tracking total resource usage. The performance metric used to track model performance was the calculated neutron efficiency at a gamma leakage of 10-3. The results of the study will be used to aid the specification of an eFPGA fabric, which will be integrated as part of a test chip.

show abstract

Section: Jinst 19 P07034mentioning

confidence: 99%

Investigating resource-efficient neutron/gamma classification ML models targeting eFPGAs

Johnson,

Boxer,

Prakash

et al. 2024

J. Inst.

View full text Add to dashboard Cite

show abstract

“…Among them, CNNs possess many excellent characteristics, such as local connectivity, parameter sharing, and adaptability. Due to the exceptional capability of the CNN method in terms of classification, it has been used to discriminate the particles in some nuclear experiments and proven to effectively improve the performance of PSD [15][16][17].…”

Section: Introductionmentioning

confidence: 99%

n/γ discrimination for CLYC detector using a one-dimensional Convolutional Neural Network

et al. 2023

View full text Add to dashboard Cite

In this work, a one-dimensional convolutional neural network (1D-CNN) is used for performing pulse shape discrimination (PSD) between the neutrons and gamma rays detected by the Cs2LiYCl6: Ce3+(CLYC) crystal. We use three different optimizers to train the CNN for comparing the effects of different optimizers on the training results. The neural network that uses the RMSProp optimizer performed the best. The accuracy of the AmBe source reaches 99.395%, and the false alarm rates (FARs) of the gamma source, i.e., 137Cs and 22Na are only 0.003% and 0.020%, separately. By the same dataset, we introduced several other methods to compare, including the classic charge integral (CI), partial charge-to-peak ratio (PCPR), decision tree (DT), support vector machine (SVM), K-nearest neighbor (KNN), and artificial neural network (ANN). Among these introduced methods, the FARs of the ANN method are better, which are 0.004% and 0.031%; however, its error is higher than that of the CNN method. A detailed discussion of the discrimination capability as a function of the sampling rate of the digitizer is also presented. We compare the performance of the CNN method and the traditional integral method under different sampling rates. The results show that even under a low sampling rate, the discrimination of the CNN method is almost unchanged, while the accuracy of the traditional integral method deteriorates rapidly. In addition, the CNN method is used for classifying more complicated particles, neutrons, gamma-rays, noise, and pile-up waveforms. The classification results show that the CNN has the ability to separate the four signals from the dataset efficiently.

show abstract

Fast neutron coincidence counter using scintillators for safeguards verification of the TRU/RE waste

Lee

Yoon

Ahn

et al. 2023

J Radioanal Nucl Chem

View full text Add to dashboard Cite

Fast neutron-gamma discrimination in organic scintillators via convolution neural network

Cited by 10 publications

References 17 publications

Investigating resource-efficient neutron/gamma classification ML models targeting eFPGAs

Investigating resource-efficient neutron/gamma classification ML models targeting eFPGAs

n/γ discrimination for CLYC detector using a one-dimensional Convolutional Neural Network

Fast neutron coincidence counter using scintillators for safeguards verification of the TRU/RE waste

Contact Info

Product

Resources

About