基于BP神经网络的高桩码头基桩损伤识别研究  

RESEARCH ON PILE DAMAGE IDENTIFICATION OF HIGH-PILED WHARF BASED ON BP NEURAL NETWORK

作　　者：郑永来[1] 肖飞[1] 潘坦博 韩雨莘 ZHENG Yong-lai;XIAO Fei;PAN Tan-bo;HAN Yu-xin(School of Civil Engineering.Tongji University,200092,Shanghai,China)

机构地区：[1]同济大学土木工程学院,上海200092

出　　处：《建筑技术》2024年第3期371-376,共6页Architecture Technology

摘　　要：针对高桩码头基桩的损伤识别问题,基于BP神经网络开展了损伤定位研究。传统损伤定位方法在识别过程中受到人为主观因素的干扰,且对于只有一阶模态数据的情况定位效果有限。为克服这些问题,构建了不受人为因素影响的损伤定位神经网络,以第三类损伤指标ULSC和δFC作为训练样本,实现了对基桩局部损伤的准确定位。在建立合理的高桩码头有限元模型的基础上,构建了基于BP神经网络的损伤定位模型,并使用ABAQUS模拟数据和实测振动信号数据进行训练和测试。实验结果表明,该神经网络模型具有较高的定位准确性和鲁棒性,在不同损伤工况和10%噪声水平下仍表现优异。This paper addresses the problem of damage identification in the foundation piles of high-pile wharves and conducts damage localization research based on the BP neural network.Traditional damage localization methods are susceptible to subjective human factors and have limited effectiveness when dealing with only first-order modal data.To overcome these issues,we construct a damage localization neural network that is not influenced by subjective factors,using the third class damage indicators ULSC andδFC as training samples to achieve accurate localization of local damages in foundation piles.Building on a reasonablefinite element model of the high-pile wharf,we construct a BP neural network-based damage localization model and train and test it using ABAQUS simulated data and actual vibration signal data.Experimental results demonstrate that the neural network model exhibits high localization accuracy and robustness,performing exceptionally well under different damage conditions and 10%noise levels.

关 键 词：BP神经网络 损伤识别 基桩损伤 健康监测 

分 类 号：TU446[建筑科学—岩土工程]