Single event effects evaluation on Convolution Neural Network in Xilinx 28nm System on Chip

Abstract Convolutional neural networks (CNN) have great exhibit excellent performance in the areas of image recognition and object detection, which can enhance the intelligence level of spacecraft. However, in aerospace, energetic particles, such as heavy ion, proton and alpha particle, can induce single event effects (SEE) that lead to malfunction of CNN, and significantly impact the reliability of CNN system. In this paper, the MNIST CNN system was constructed based on a 28 nm System-on-Chip (SoC), and then the alpha particle irradiation experiment and fault injection were applied to evaluate the SEE of the CNN system. Various types of soft errors in the CNN system have been detected, and the SEE cross sections have been calculated. Furthermore, the mechanisms behind some soft errors have been explained. The research will provide technical support for designing of radiation-resistant artificial intelligence chips.