基于IPSO-BP神经网络的管道损伤检测方法  

The Method of Pipeline Damage Detection Based on IPSO-BP Neural Network

作　　者：吴磊 梅江涛 赵硕 WU Lei;MEI Jiangtao;ZHAO Shuo(School of Petroleum Engineering,China University of Petroleum(East China),Qingdao 266580,Shandong,China)

机构地区：[1]中国石油大学(华东)石油工程学院,山东青岛266580

出　　处：《实验室研究与探索》2022年第7期44-49,共6页Research and Exploration In Laboratory

基　　金：山东省泰山学者科研项目(tsqn201909067);山东省自然科学基金项目(ZR2020QE300)。

摘　　要：为了更高效准确地对流质输送管道进行损伤检测,结合改进粒子群算法和误差反向传播神经网络,提出了一种IPSO-BP神经网络模型,并以IPSO-BP网络和IPSO算法建立损伤位置和损伤程度识别器。针对管道损伤检测问题,提出以曲率模态和位移模态为输入特征参数,由损伤位置识别器、损伤程度识别器构成的管道结构损伤识别模型。基于管道的特定工况,对不同损伤状态下的管道进行模态分析和静力学分析,得到共100组训练样本和12组测试样本,利用IPSO-BP神经网络进行结构损伤识别模型的训练和测试。结果表明:提出的基于IPSO-BP神经网络的管道损伤识别模型对损伤位置的识别准确率为100%,损伤程度误差率低于5%,该模型为各种工况下的管道损伤检测提供了一类快速准确的方法。In this study,by combining improved IPSO with BPNN,an IPSO-BP neural network model is proposed,and the damage location and damage degree recognizersare established based on IPSO-BP network and IPSO algorithm.Aiming at detecting of pipeline damage,a pipeline structure damage identification model is proposed,which takes curvature mode and displacement mode as the input characteristic parameters,and consists of damage location identifier and damage degree identifier.Based on the specific working conditions of pipeline,the modal analysis and static analysis under different damage states are carried out,and a total of 100 groups of training samples and 12 groups of test samples are obtained.The IPSO-BP neural network is used to train and test the structural damage identification model.The experimental results show that the accuracy of the pipeline damage identification model based on IPSO-BP neural network is 100%,and the error rate of damage degree is less than 5%.The study provides a fast and accurate method for pipeline damage detection under various working conditions.

关 键 词：粒子群优化算法 BP神经网络 管道损伤检测 曲率模态 位移模态 

分 类 号：U178[交通运输工程]