基于神经网络的多过失误差侦破方法的研究

基于神经网络的过失误差侦破方法具有简单、计算量小和适于在线应用的优点，并且相对于传统方法具有处理非线性问题能力较强的特点。但是在侦破多过失误差时，现有的直接侦破法和序列侦破法的侦破率较低。针对这一情况，本文提出了将神经网络和测量数据检验法相结合的侦破多过失误差的新方法，该方法首先利用神经网络较强的鲁棒性和容错能力对数据进行处理，然后再进行过失误差侦破。实例研究表明，这种方法能够有效地提高多过失误差共存时的侦破能力。

Studies on detecting multiple simultaneous gross errors via artificial neural networks

The methods of detecting gross errors in measured data based on neural networks have the advantages of simple and fittingonline use.At present the main methods for multiple gross errors detection are straight detection method and serial detection method.But those methods have low detection ratio when applied to multiple gross errors detection.This paper presents a new method which isthe combination of neural networks method and measurement test method.The new method first utilizes the neural networks to processthe measured data;then detects gross errors by statistical method.The comparison between the new method and the serial detectionmethod are made by an example.The results show that the new method can improve the detecting power of the muhiple gross errors de-tection greatly.