中文摘要：

针对模拟电路故障的实时诊断问题,提出使用嵌入式系统实现BP神经网络进行诊断的解决方案。对诊断系统的芯片选择、系统架构、诊断流程、训练算法等问题进行了研究,提出使用STM32芯片作为核心芯片以及加入动量项的反向传播算法进行训练的策略,明确了系统的结构及运行流程。最后使用设计的系统分别进行了线性和非线性模拟电路的故障诊断实验进行验证。实验结果表明,基于嵌入式系统的BP神经网络能够有效及时地诊断出模拟电路的故障。

英文摘要：

BP neural network based on embedded system was proposed for diagnosing fault analog circuits. The problem in the chip selection, system architecture, diagnosis processes, training algorithms was studied. STM32 was chosen as the core chip and the added momentum of back-propagation algorithm was used as training strategy, the structure and operation of process system was determined. The experiments were carried out for validation of fault diagnosis of linear and nonlinear analog circuits of the designed systems. The results showed that the BP neural network based on embedded system could diagnose analog circuit fault effectively and timely.