模拟电路故障诊断的BP神经网络方法研究

BP网络是模拟电路故障诊断中应用十分广泛的一种神经网络。针对传统BP算法的误差下降缓慢，调整时间长，甚至容易陷入局部极小点而不能自拔等局限性，提出用弹性算法与BP网络相结合的方法，并结合某型雷达装备的具体电路，运用该方法建模、仿真。实验结果表明，采用弹性算法结合后的BP网络误差收敛稳定，训练速度快，在克服传统BP算法的局限性上效果显著，为新型雷达装备的故障诊断和维修提供了一种方法，具有特定的实用意义。

The BP network is a neural network which is widely used in fault diagnosis for analog circuits. Due to the dis- advantages of the BP traditional algorithm that the error is dropped slowly, the adjustment time is long, and even is easily to fall into local minimum and unable to extricate themselves and so on, a method of combining resilient arithmetic with BP net- work is proposed. Then the modeling and simulation are performed with a certain type of radar equipment in the specific cir- cuit. Experimental results show that the error of BP network which combining with resilient arithmetic has stabile convergence and fast training. And the effect in overcoming the limitations of traditional BP algorithms is obvious. A method of fault diag- ...