基于BP神经网络的风机转子故障诊断

Fault diagnosis of rotor in fan based on BP NN

应用BP算法对转子进行故障诊断 ,首先论述了BP算法的基本原理 ,然后提出了输出模式矩阵的改进方法 ,即将常用的单位矩阵来表示的故障模式转变为二进制结构 ,由此简化了网络结构 ,明显减少了样本训练次数 ,提高了计算故障诊断中对单一故障尤为有效 .最后通过实例进行分析 ,证明这一方法对提高运算速度十分有效 .

In this paper, the BP arithmetic is applied to the fault diagnosis of the rotor. At first, the fundamental of BP arithmetic is related, and then the ameliorated method of output mode matrix is put forward. It transferred the identity matrix in common use into the binary system.It simplified the structure of the network and therefore reduced the number of the training swatches, so the calculating efficiency is improved. In practice, the algorithm is very efficient in the fault diagnosis of the complicated rotor machine. Finally,the efficiency of the method in improment of calculating speed is illustrated.