摘要
故障诊断对于事故后快速恢复具有重要的意义。模拟电路故障诊断有许多方法,提出了一种基于遗传算法优化的BP神经网络智能诊断技术。该方法采用基于实数编码的遗传算法优化神经网络权值和阈值,代替了原来BP网络随机设定的初始权值和阈值。然后再用改进的BP算法用已由遗传算法确定的空间对网络进行精确搜索。实验仿真结果表明基于遗传算法优化过的神经网络的训练步数得到大大的减少,泛化能力也得到提高。克服了传统BP算法的收敛速度慢,容易陷入局部极小的缺点。
Fault diagnosis is of great importance to quick recovery. This paper presents an intelligent diagnosis method based on genetic algorithm（GA） optimized BP Neural Networks. The method adopts an improved genetic algorithm based on real - coding. The simulation results indicate that the genetic algorithm （GA） optimized BP Neural Networks tran＇ s epoch decrease, and generalization ability has been enhanced. It overcomes the shortcomings of traditional BPNN, such as low convergence speed, easy to fall into the local minimum points.
出处
《计算机仿真》
CSCD
2007年第9期293-296,共4页
Computer Simulation
关键词
故障诊断
神经网络
遗传算法
模拟电路
Fault diagnosis
Neural networks
Genetic algorithm
Analog circuit
