基于BP神经网络的模拟电路故障诊断

Fault Diagnosis of Analog Circuit Based on BP Neural Network

模拟电路的可靠性决定了电子设备的可靠性。针对目前BP网络在模拟电路故障诊断中存在的容差和非线性等问题,提出多频组合法,对电路中的软故障进行诊断。通过在电路输入端施加3 V的正弦激励,选取不同的测试频率,对测试点进行测试,得到各待测元件的故障值。该系统解决了模拟电路中的容差问题,提高了诊断率,并在仿真中得以证明。

The analog fault diagnosis research is one of the forefront of the testing area, one of the reason is that the reliability of electrical systems depends on that of analog circuits and systems. In accordance with the shortcoming that BP neural network exists, such as tolerance, nonlinear problem and other problems, there is a kind of multi - frequency combination method. The method establishes the analog diagnosis of software. Through exerting 3 V Sine drive in the input end of the circuit, selecting the different test frequency to the test point, it can obtain the breakdown value of each testing part. The primary aim of the paper is to provide a mechanism to deal with the problem of element tolerance and reduce the testing time. The simulation results show that the proposed method can perform correct diagnosis in the analog circuits with tolerance.