摘要
将模糊逻辑与神经网络相结合,构造模糊神经网络,将神经网络输入层的确定性信息模糊化后变成模糊量,将故障征兆参数相对应的隶属度数值作为神经网络的输入,从而使神经网络更加适合设备故障描述,克服了神经网络对不精确信息表达的缺点。提出基于黄金分割法的变步长BP算法来训练神经网络,根据误差变化趋势动态调整学习速率,实现学习步长的自适应调整,提高网络收敛的速度,防止网络训练时陷入局部极小。将训练好的模糊神经网络应用于抽油机设备的故障诊断,取得良好效果。
Combining with fuzzy logic and neural network,fuzzy neural network is constructed.After the neural network layer of certainty input information turns into fuzzy sense then becomes fuzzy quantity.Take the fault symptom parameters corresponding to the membership values as the input of network,then the neural network is more suitable for equipment fault description,it overcomes the neural network defect of inaccurate information expression.Based on the golden section method of variable step-size BP algorithm to train the neural network.According to the error tendency dynamically adjusting rate,achieves the learning step of adaptive adjustment.It improves the network convergence speed,prevents network training sink into local minimum.The trained fuzzy neural network is applied to fault diagnosis of pumping equipments and achieves good results.
出处
《自动化技术与应用》
2012年第4期4-8,共5页
Techniques of Automation and Applications
基金
黑龙江省教育厅科学技术研究项目资助(编号12511014)
关键词
模糊神经网络
BP算法
黄金分割法
抽油机
故障诊断
fuzzy neural network
BP algorithm
golden section method
pumping unit
fault diagnosis