A 1-D CNN Algorithm for Low-Background β Detection With Time Projection Chamber

Abstract

Low-background $\beta $ detection is crucial for environmental safety. In this article, a 1-D convolutional neural network (1-D CNN)-based algorithm is introduced for low-background $\mathrm {\beta }$ detection using time projection chambers (TPCs), aiming to classify $\mathrm {\beta }$ and background signals captured by the detectors and recorded using the electronic system. Experimental results demonstrate the effectiveness of the proposed algorithm in handling complex background and $\mathrm {\beta }$ signals. The neural network was trained and tested on two datasets from different conditions. In each dataset, the test result showed a background rejection rate of over 98%, with a $\mathrm {\beta }$ retention rate of approximately 55%. Compared to traditional lead-shielded detection methods, the application of this algorithm enables lead-free, low-background $\beta $ detection. This can lead to a significant reduction in instrument size and weight, thereby greatly expanding its potential applications.