基于证据理论的装备损伤定位方法中的关键问题研究

摘要

To solve the problem of rapid positioning problem of equipment damage ,the damage location method is put forward based on D-S evidence theory .This method is to unit based on the damage location ,damage tree combination of evi-dence theory and BP neural network ,through the establishment of damage events corresponding to the damage location of ev-idence collection to locate the damage .In order to solve the problem of evidence theory needs to be known a priori probabili-ty ,by using BP neural network will be the usual maintenance effective information and expert experience ,the combination of training a new probability distribution ,and good evidence is set up into the body in the collection ,and higher precision of damage location information .Application of this method is the key problem of the transformation of the feature extraction and evidence of damage .%为解决装备损伤的快速定位问题，论文提出了基于 D-S 证据理论的损伤定位方法。该方法以损伤树单位置损伤定位为基础，将证据理论与 BP 神经网络结合，通过建立损伤事件对应损伤位置的证据集合实现多位置损伤定位。为了解决证据理论需要已知先验概率的问题，利用 BP 神经网络将平时维修得到的有效信息和专家经验相结合，训练确定出新的概率分配，并融入到建立好的证据体集合中，从而更高精准度地损伤定位信息。应用该方法的关键问题是损伤特征的提取和证据的转化。