改进的BP网络及其在电路故障诊断中的应用

针对标准BP算法收敛速度慢、容易陷入局部极小值的缺点，提出了改进的BP算法——Levenberg-Marquardt算法，并将其应用到电路故障诊断问题中。实例仿真验证了改进算法在训练次数和训练时间上具有明显的优势，同时也能保证故障诊断准确率，具有良好的效果。

Application of an Improved BP Network in Circuit Fault Diagnosis

BP neural network has been applied in many fidds. However, its application is limited by its own defection which includes that BP algorithm is standstill in learning and the state of BP network is subject to getting trapped in one of local minima. An improved algorithm - Levenberg- Marquardt algorithm is presented in this paper and it is applied to solve such a problem of circuit fault diagnosis. An illustrative example is given to validate that the improved one is advantageous at the aspects of epoch and elapsed time over the traditional one and the accuracy of diagnosis is kept and its performance is well.