摘要
BP神经网络在船舶发电机故障诊断中有广泛的应用,但由于BP网络采用的是沿梯度下降的搜索求解算法,存在收敛速度慢、且容易陷入局部极小等问题,给故障诊断带来不便。为此,采用蚁群优化算法代替反向传播算法训练神经网络的权值和阈值。以船舶发电机中的同步发电机为例,利用训练后的蚁群神经网络对其进行故障诊断,并把BP神经网络和蚁群神经网络的训练和诊断结果相比较,结果表明蚁群神经网络具有较好的训练性能、收敛速度、诊断精度和良好的故障识别率,应用于船舶发电机的故障诊断中,具有较好的应用前景。
BP neural network are widely used in the fault diagnosis of ship generator.But as BP neural network adopts search algorithm of the gradient descent,which has the problem of slow convergence rate and easily into the local minimum,bring inconvenience to malfunction diagnose.Therefore adopting ant colony optimization algorithm to take the place of propagation algorithm to train the neural network weights and thresholds.With an example of synchronous generator marine generator,utilizing the trained neural network for malfunction diagnosis of marine generator,and comparing with the BP algorithm and the result of genetic neural network training and diagnosis,the results show that the neural network which is based on the ant colony optimization algorithm,has a better training performance,convergence speed,accuracy and good fault diagnosis recognition rate.It will have a good prospect of application in the fault diagnosis of generator.
出处
《科学技术与工程》
2010年第22期5595-5598,共4页
Science Technology and Engineering
关键词
船舶发电机
故障诊断
蚁群算法
神经网络
ship generator fault diagnosis ant colony algorithm neural network
