改进遗传算法优化的神经网络在智能故障诊断中的应用

Application of the genetic simulated annealing algorithm based on neural network in failure diagnosis

设计了用模拟退火的混合遗传算法代替BP网络的反向传播过程的改进算法,解决了在机械故障诊断系统中BP算法容易陷入局部极小值的问题。该算法是在遗传法中引入模拟退火机制,将其同BP算法结合,形成一个混合的优化算法。新算法既有神经网络的学习能力和鲁棒性,又有遗传算法的强的全局随机搜索能力。仿真结果表明,这种改进算法极大提高了内燃机故障诊断系统的效率和准确性。

The improved arithmetic designed is substituted for the backward propagation process of BP arithmetic. This optimization BP algorithm introduces simulated annealing algorithm to genetic algorithm and have it combined with multilayer perceptrons to form a compounding optimization perceptron. The new perceptron have both the learning ability of the neural network and robustness,and strong random searching ability of genetic algorithm, h isshown by simulated results that the improved arithmetic can greatly improve efficiency and accuracy of Internal Combustion Engine failure diagnosis system.