基于遗传粒子群算法的模拟电路故障诊断方法研究

Analog Circuit Fault Diagnosis Based on Benetic Particle Swarm Optimization Algorithm(GAPSO)

针对模拟电路故障模式复杂多样的特点,提出一种基于小波包分解、归一化处理、遗传粒子群优化算法(GAPSO)和BP神经网络相结合的模拟电路故障诊断新方法;该方法使用小波包分解重构获取各尺度函数空间上的能量特征信息作为特征向量输入神经网络进行训练和诊断;利用遗传粒子群算法优化BP神经网络的权值和阈值,能有效克服BP神经网络极易陷入局部极小等缺陷;通过Multisim仿真电路实例,比较GAPSO-BP和BP神经网络的诊断结果,得知GAPSO-BP神经网络能够较为有效地实现模拟电路的故障诊断,具有一定的实际应用价值。

This paper puts forward a new method of analog circuit fault diagnosis based on the combination of wavelet packet decomposi- tion, the normalized processing, GAPSO and BP neural network for the analog circuit mode of complex and diverse characteristics. The method, adopting wavelet packet decomposition, obtains each scale energy characteristics information from function space as its vector to put in the neural network. It uses GAPSO to optimize the weights and thresholds of BP, which can effectively overcome the defects of BP neural network such as easy to fall into local minimum. Taking Multisim simulation circuit as an example, the effectiveness of the proposed method is verified by the comparing of the GAPSO--BP and BP result. GAPSO--BP neural network can be more effectively achieved analog circuit fault diagnosis and has a certain practical application value.