基于BP神经网络的复杂布局管道泄漏检测研究  

Research on Complicated Pipeline Leakage Detection Based on BP Neural Network

作　　者：阳子轩[1] 范世东[1] 江攀[1] 熊庭[1] 

机构地区：[1]武汉理工大学能源与动力工程学院,武汉430063

出　　处：《船海工程》2012年第1期66-69,共4页Ship & Ocean Engineering

基　　金：高等学校学科创新引智计划(B08031)

摘　　要：对基于BP神经网络的复杂布局管道泄漏检测与定位进行研究,确定实验方案和研究内容,介绍BP神经网络的输入参数、训练目标和隐含层节点个数,阐述基于BP神经网络的复杂管道泄漏检测研究。研究表明:用BP神经网络来预测复杂管道发生泄漏后的泄漏量大小是完全可行的;用其对是否发生泄漏进行状态辨识,则还要进行数据的二次处理;由于训练样本过少,用BP网络进行泄漏定位时会产生较大的误差,建议增加训练样本数,从而提高泄漏定位精度。The leakage detection and location of complicate pipeline was studies based on BP neural network. After defining the experimental program and research content, the input parameters, training objective and the number of hidden nodes of BP neural network were introduced. The research showed that it is feasible to predict the leakage in the complicated pipeline. When use BP neural network to identify if there have leakage in the pipeline, the data needs to be secondary processed. As training sample is few, it may causes deviation in the leakage location. But increasing the number of training sample could improve the leakage location accuracy.

关 键 词：BP神经网络 复杂管道 泄漏检测 

分 类 号：U671.97[交通运输工程—船舶及航道工程]