基于忆阻器神经网络的辐射效应诊断系统设计

Design of Diagnosis System of Radiation Effect Based on Memristor Neural Network

航天事业是我国群策群力发展的重要事业,航天器作为执行航天任务的主体,其可靠性至关重要。随着集成电路工艺技术和人工智能的迅速发展,使用传统集成电路实现神经网络将会耗费大量的硬件资源,忆阻器是一种可与传统集成电路集成的新型器件且可以用于神经网络计算。针对集成电路中静态随机存取存储器的辐射故障问题,设计了一种卷积神经网络,对集成电路辐射效应进行诊断,基于忆阻器对神经网络的运算层进行设计,通过设计的卷积神经网络对集成电路的故障数据集实现了分类诊断。

The aerospace industry has always been an important cause of China’s vigorous development.As a main body performing space missions,its reliability is crucial.The performance of integrated circuits and artificial intelligence are developing rapidly,but the use of traditional integrated circuits to implement neural networks will consume a lot of hardware resources.Memristors are a new type of device that can be integrated with traditional integrated circuits and can be used for neural network computing.Aiming at the problem of radiation fault of static random access memory integrated circuits,a convolutional neural network is designed to diagnose the radiation effect of integrated circuits.The operation layer of the neural network is designed based on memristors,and the fault data set of the integrated circuit is designed through the designed convolutional neural network to realize classification diagnosis.