利用SOM网络可视化方法诊断凝汽器系统故障

Study on Fault Diagnosis of Steam Condenser Based on SOM Visualization Neural Network

SOM网络是一种重要的无导师学习训练算法的神经网络,使用该算法进行训练后,可以将高维输入空间映射到二维空间上,并对故障现象进行自动分类,从而得出它们对应的故障原因。同时为了便于识别故障,利用U矩阵图对神经网络仿真结果进行可视化。以凝汽器故障诊断的为例,总结了凝汽器的故障集、征兆集和故障特征数据。利用SOM网络进行仿真并对其结果进行可视化分析,表明SOM网络可视化方法具有故障模式识别能力,是一种可行有效的凝汽器故障诊断方法。

SOM network is an important non-learning mentors training arithmetic of neural network After arithmetic come into used, the high-dimensional space is mapped to the two-dimensional space, and failure phenomenon carry out automatic classification, accordingly finding failure reasons for be distinguished as failure, There is make neural network of simulation result to visualization by U matrix. Take example for of the condenser sets, the paper includes failure, signs and failure features date of condenser sets. Simulation resuh maked visualization analysis by SOM neural network that be provided with failure mode of discriminating capability, this capability is a feasible and effective failure diagnosis method.