在多参数检测时神经网络的结构优化方法

Method for optimizing structure of neural network in states measuring based on multi-parameters

运用多特征参数进行设备状态监测是一种具有较高精确度的技术手段,但是当检测参数数量较大时运算量是一个不容忽视的问题。针对采用人工神经网络进行设备状态检测时的结构优化问题进行研究,提出由特征压缩层和检测层组成串联网络的方法建立神经网络检测模型。仿真结果表明,该组合网络减小了运算量,改善了网络收敛性能。

It is an effective method to measure the states of a device using multi-parameters,but the calculations are considerable when the number of the measured parameters is large,so such a problem should not be neglected.This paper studies the structure optimization of artificial neural network when measuring the state of a device,puts forward a combined neural network model including the functions of compressing characteristic, and diagnosing faults.The result after simulating shows that such kind of combined network reduces the calculations and also improves the performance of convergence.