Convolutional Neural Networks (CNNs) are currently one of the most widely used predictive models in machine learning. Recent studies have demonstrated that hardware faults induced by radiation fields, including cosmic rays, may significantly impact the CNN inference leading to wrong predictions. Therefore, ensuring the reliability of CNNs is crucial, especially for safety-critical systems. In the literature, several works propose reliability assessments of CNNs mainly based on statistically injected faults. This work presents a software emulator capable of injecting real faults retrieved from radiation tests. Specifically, from the device characterisation of a DRAM memory, we extracted event rates and fault models. The software emulator can reproduce their incidence and access their effect on CNN applications with a reliability assessment precision close to the physical one. Radiation-based physical injections and emulator-based injections are performed on three CNNs (LeNet-5) exploiting different data representations. Their outcomes are compared, and the software results evidence that the emulator is able to reproduce the faulty behaviours observed during the radiation tests for the targeted CNNs. This approach leads to a more concise use of radiation experiments since the extracted fault models can be reused to explore different scenarios (e.g., impact on a different application).

Accelerated terrestrial neutron irradiations were performed on different commercial SiC power MOSFETs with planar, trench, and double-trench architectures. The results were used to calculate the failure cross sections and the failure-in-time (FIT) rates at sea level. Enhanced gate and drain leakage were observed in some devices which did not exhibit a destructive failure during the exposure. In particular, a different mechanism was observed for planar and trench gate MOSFETs, the first showing a partial gate rupture with a leakage path mostly between the drain and the gate, similar to what was previously observed with heavy ions, while the second exhibiting a complete gate rupture. The observed failure mechanisms and the postirradiation gate stress (PIGS) tests are discussed for different technologies.

In this article, we investigate the reliability of Google’s coral tensor processing units (TPUs) to both high-energy atmospheric neutrons (at ChipIR) and thermal neutrons from a pulsed source [at equipment materials and mechanics analyzer (EMMA)] and from a reactor [at Thermal and Epithermal Neutron Irradiation Station (TENIS)]. We report data obtained with an overall fluence of $3.41\times {10^{12}} {\mathrm {n/cm}}^{2}$ for atmospheric neutrons (equivalent to more than 30 million years of natural irradiation) and of $7.55\times {10^{12}} {\mathrm {n/cm}}^{2}$ for thermal neutrons. We evaluate the behavior of TPUs executing elementary operations with increasing input sizes (standard convolutions or depthwise convolutions) as well as eight convolutional neural networks (CNNs) configurations (single-shot multibox detection (SSD) MobileNet v2 and SSD MobileDet, trained with COCO dataset, and Inception v4 and ResNet-50, with ILSVRC2012 dataset). We found that, despite the high error rate, most neutron-induced errors only slightly modify the convolution output and do not change the detection or classification of CNNs. By reporting details about the error model, we provide valuable information on how to design the CNNs to avoid neutron-induced events to lead to misdetections or classifications.

This paper studies the impact of radiation-induced Single Event Effects (SEEs) in the Zynq-7000 FPGA and presents an in-depth analysis of the SEE susceptibility of all the memories of the programmable logic. The radiation experiments were per-formed in the CERN North Area facility and in the GSI Helmholtz Centre for Heavy Ion Research using very/ultra-high energy heavy ions. The offline analysis of the radiation experimental results pro-duced a deep understanding for various SEE phenomena observed in the Zynq-7000 FPGAs, such as SEFIs, SET in global signals and multiple bit upsets that could be key issues for the design of an effective SEE mitigation approach.

EdgeAI is an emerging AI accelerator technology which is capable of delivering improved AI performance at both a lower cost and a lower power level. With the aim of implementation in large quantities and in safety-critical environments, it is imperative to understand how Single Event Effects (SEE) affect the reliability of this new family of devices, and to propose efficient hardening solutions. Through neutron beam experiments and fault-injection analysis of a COTS EdgeAI device, we are able to identify the device’s SEE failure-modes, separate the error rate contributions of the device’s different resources, and characterise the device’s SEE reliability. During this analysis we discovered that the vast majority of single bit flips have no appreciable effect on the output. After this analysis, we propose a hardening solution which implements TMR in the device without changing its physical architecture. We experimentally validate this solution and show that we are able to correct 96% of the misclassifications (critical errors) with nearly-zero overhead.