Artificial neural network algorithm for pulse shape discrimination in 2πα and 2πβ particle surface emission rate measurements

• Published in: Nuclear science and techniques Vol. 34; no. 10; pp. 91 - 102
• Main Authors: Li, Yuan-Qiao, Zhu, Bao-Ji, Lv, Yang, Zhu, Heng, Lin, Min, Chen, Ke-Sheng, Xu, Li-Jun
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.10.2023; China Institute of Atomic Energy,Beijing 102413,China%China Institute of Atomic Energy,Beijing 102413,China; National Key Laboratory for Metrology and Calibration Techniques,Beijing 102413,China
• Subjects: Beam Physics; Nuclear Energy; Particle Acceleration and Detection; Particle and Nuclear Physics; Physics; Physics and Astronomy; Multi-wire proportional counter; Alpha and beta sources; Pulse shape discrimination; Surface emission rate; Artificial neural networks
• ISSN: 1001-8042; 2210-3147
• DOI: 10.1007/s41365-023-01305-0

To enhance the accuracy of 2π α and 2π β particle surface emission rate measurements and address the identification issues of nuclides in conventional methods, this study introduces two artificial neural network (ANN) algorithms: back-propagation (BP) and genetic algorithm-based back-propagation (GA-BP). These algorithms classify pulse signals from distinct α and β particles. Their discrimination efficacy is assessed by simulating standard pulse signals and those produced by contaminated sources, mixing α and β particles within the detector. This study initially showcases energy spectrum measurement outcomes, subsequently tests the ANNs on the measurement and validation datasets, and contrasts the pulse shape discrimination efficacy of both algorithms. Experimental findings reveal that the proportional counter’s energy resolution is not ideal, thus rendering energy analysis insufficient for distinguishing between 2π α and 2π β particles. The BP neural network realizes approximately 99% accuracy for 2π α particles and approximately 95% for 2π β particles, thus surpassing the GA-BP’s performance. Additionally, the results suggest enhancing β particle discrimination accuracy by increasing the digital acquisition card’s threshold lower limit. This study offers an advanced solution for the 2π α and 2π β surface emission rate measurement method, presenting superior adaptability and scalability over conventional techniques.