摘要
轴心轨迹的图形形状识别是旋转机械故障诊断中最为重要的内容之一。利用图形的不变性特征矩的识别技术对轴心轨迹的图形数据进行特征提取,并将特征提取结果作为神经网络的输入让其学习,学习完成后,就可利用神经网络对轴心轨迹的图形形状进行分类识别。神经网络采用改进的BP网络,提出的一种确定神经网络最优隐含层节点数的新方法,其正确性得到了大量事实的验证。对仿真轴心轨迹图形形状的识别结果表明,该方法是有效的、可行的。
The identification of the graphic shape of core path is one of the most important contents in the failure diagnosis of rotation machinery. The characteristics of graphic data of core path is collected by the identification technology of graphic invariance characteristic matrix and then the collected characteristic results is input into the neural net to let it learn. After the completion of learning, the neural net can be used to make the classification identification on the graphic shape of core path. The neural net adopts the improved BP net, a new method of determining the node number of optimum implication layer of neural net, and its correctness has been proved by a lot of facts. The results of the identification of graphic shape of simulation core path showed this method is effective and feasible.
出处
《水力发电》
北大核心
2002年第12期34-37,共4页
Water Power
