基于EfficientNetV2的模拟电路早期软故障诊断

Early Soft Fault Diagnosis in Analog Circuits Based on EfficientNetV2

摘要:

针对模拟电路早期软故障诊断困难问题,提出了一种基于卷积神经网络EfficientNetV2的模拟电路早期软故障诊断方法,该方法使用扫描信号作为被测电路的激励信号,采集被测电路输出端基于各种软故障的原始信号,利用连续小波变换进行时频分析,将输出的时域故障信号转化为二维时频图,作为EfficientNetV2网络的输入,利用EfficientNetV2网络提取模拟电路的故障特征,确定元件的故障类型,实现电路的故障诊断,用Sallen-Key带通滤波电路和四运放双二次滤波电路进行了仿真实验,实验结果表明,该方法在测试电路上表现优异,测试电路的诊断准确率均高达99.6%。

Abstract:

In order to solve the problem of early soft fault diagnosis in analog circuits, a method based on convolutional neural network EfficientNetV2 for early soft fault diagnosis in analog circuits is proposed. This method uses scanning signals as excitation signals for the circuit under test, collects raw signals based on various soft faults at the output of the circuit under test, performs time-frequency analysis using continuous wavelet transform, converts the output time-domain fault signal into a two-dimensional time-frequency map, which is used as the input of EfficientNetV2 network. The EfficientNetV2 network is used to extract fault features of the analog circuit to determine the fault type of the component and achieve fault diagnosis of the circuit. Simulation experiments are conducted using Sallen-Key bandpass filter circuit and four op-amp dual quadratic filter circuit. The experimental results show that this method performs well on the test circuit, and the diagnostic accuracy of the test circuit is as high as 99.6%.