南京工业大学学报（自然科学版）2025，Vol.47Issue(5)：532-542,11.DOI:10.3969/j.issn.1671-7627.2025.05.007
基于卷积神经网络-双向长短期记忆网络模型的黏滞阻尼器监测数据降噪与损伤预警
CNN-BiLSTM based approach for denoising and damage early warning in the data from viscous damper monitoring
摘要
Abstract
Viscous dampers effectively reduce structural damage during earthquakes by capturing and dissipating seismic energy.However,it is worth noting that if these dampers have sustained damage prior to seismic events,their protective performance is significantly compromised.Therefore,continuous monitoring of the operational condition of viscous dampers and timely identification of potential damage are essential.This paper first preprocesses long-term monitoring data of viscous dampers and proposes a novel denoising strategy that integrates Empirical Mode Decomposition(EMD)with Discrete Wavelet Transform(DWT).The denoised signal achieves a signal-to-noise ratio(SNR)of 46.387 dB,representing an 11.3％improvement over the use of EMD alone.Subsequently,a Convolutional Neural Network-Bidirectional Long Short-Term Memory(CNN-BiLSTM)model is utilized to learn and predict the oil pressure data of the dampers under normal operating conditions,thereby extracting inherent characteristics and patterns within the data representing the healthy state of the damper.Finally,predictions and early warnings are performed on both oil pressure data under routine operating conditions and simulated displacement data under seismic loading to validate the effectiveness of the proposed method.Results indicate that the combined EMD-DWT denoising method effectively corrects distorted monitoring data from viscous dampers and that the CNN-BiLSTM based prediction and early warning method effectively detects potential damage in viscous dampers.The method proposed in this study plays a crucial role in ensuring the safe and stable operation of viscous dampers and their associated structures,providing engineering guidance for viscous damper monitoring and early warning systems.关键词
黏滞阻尼器/结构健康监测/机器学习/CNN-BiLSTM模型/损伤预警/数据降噪Key words
viscous damper/structural health monitoring/machine learning/CNN-BiLSTM model/damage early warning/data denoising分类
建筑与水利引用本文复制引用
马雨贤,石晟,王曙光,耿岩,杜东升,杨伟..基于卷积神经网络-双向长短期记忆网络模型的黏滞阻尼器监测数据降噪与损伤预警[J].南京工业大学学报（自然科学版）,2025,47(5):532-542,11.基金项目
江苏省自然科学基金(BK20241074) （BK20241074）
江苏省高等学校基础科学(自然科学)研究重大项目(22KJA560003) （自然科学）
中冶集团重大项目(YCC2021Kt01) （YCC2021Kt01）
