摘要
传统经验模态分解(empirical mode decomposition, EMD)方法在处理桥梁挠度信号时存在模态混叠、分解误差累积等问题,致使分解结果尚不理想。为此,提出了一种结合变分模态分解(variational mode decomposition, VMD)和K-L散度(Kullback-Leibler divergence, KLD)的桥梁挠度监测数据温度效应分离方法。利用VMD分解桥梁挠度信号,获得若干个本征模态函数(intrinsic mode function, IMF);借助核密度估计求得各IMF分量的概率密度函数分布,进而得到各分量KLD值,剔除虚假IMF分量,选定最佳分量;运用Pearson相关系数对最佳分量进行效果评价;通过数值仿真算例与实桥监测数据,验证了该方法的有效性。结果表明:该方法融合了VMD自适应、抗噪能力强和KLD快速选取最优信号的优势,克服了传统EMD模态混叠等缺陷,减少了虚假分量的干扰,将两者结合使得分解及筛选特征信号分量高效可靠,温度效应分离效果良好;仿真信号经VMD-KLD分析得到日、年温差效应及长期挠度相关系数分别为0.994 6、0.983 7和0.970 4,实测信号得到的日、年温差效应相关系数分别为0.908 1、0.936 4;同EMD-KLD相比,VMD-KLD分离出的各挠度成分相关系数更接近于1,仿真信号分析中日、年温差效应及长期挠度分别提升了4.43%、10.84%和8.81%,实测信号分析中日、年温差效应分别提升了12.35%、5.57%。该方法可为桥梁挠度监测数据温度效应在线分离提供一种新的思路。
The traditional empirical mode decomposition(EMD) method has some problems in processing bridge deflection signals, such as, modes aliasing and decomposition error accumulation to generate unsatisfactory decomposition results. Here, a temperature effect separation method of bridge deflection monitoring data based on variational mode decomposition(VMD) and Kullback-Leibler divergence(KLD) was proposed. Firstly, the bridge deflection signal was decomposed with VMD to obtain several intrinsic mode functions(IMFs). Secondly, the probability density function of each IMF component was obtained using the kernel density estimation, and then the KLD value of each IMF component was obtained. The false IMF component was eliminated and the optimal one was selected. Thirdly, Pearson correlation coefficient was used to do the effect evaluation of the optimal component. Finally, the effectiveness of the proposed method was verified using numerical simulation examples and actual bridge monitoring data. The results showed that: the proposed method fuses VMD’s advantages of self-adaptive and strong anti-noise ability, and KLD’s advantages of rapidly choosing the optimal signal to overcome the traditional EMD’s defects of modes aliasing, etc., and reduce interference of false components;the proposed method combines VMD and KLD to make decomposing and screening characteristic signal components efficient and reliable with good temperature effect separation effect;3 correlation coefficients of daily temperature difference effect, annual temperature difference effect and long-term deflection obtained using VMD-KLD for simulation signals are 0.994 6, 0.983 7 and 0.970 4, respectively, and 2 correlation coefficients of daily temperature difference effect and annual temperature difference effect for actually measured signals are 0.908 1 and 0.936 4, respectively;compared with the results obtained using EMD-KLD, the correlation coefficient of each deflection component separated using VMD-KLD is closer to 1, 3 correlation coefficients of daily temperature difference effect, annual temperature difference effect and long-term deflection in simulation signal analysis using VMD-KLD increase by 4.43%, 10.84% and 8.81%, respectively, and 2 correlation coefficients of daily temperature difference effect and annual temperature difference effect in actually measured signal analysis using VMD-KLD increase by 12.35% and 5.57%, respectively;the proposed method can provide a new idea for temperature effect on-line separation of bridge deflection monitoring data.
作者
李双江
辛景舟
付雷
唐启智
赵月明
周建庭
LI Shuangjiang;XIN Jingzhou;FU Lei;TANG Qizhi;ZHAO Yueming;ZHOU Jianting(State Key Lab of Mountain Bridge and Tunnel Engineering,Chongqing Jiaotong University,Chongqing 400074,China;Guangxi Communications Investment Group Co.,Ltd.,Nanning 530000,China;Guizhou Bridge Construction Group Co.,Ltd.,Guiyang 550000,China;Guizhou Bijie High Speed Development Co.,Ltd.,Bijie 551700,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2022年第5期105-113,共9页
Journal of Vibration and Shock
基金
国家自然科学基金(51978111,51908094)
重庆市自然科学基金创新群体科学基金(cstc2019jcyj-cxttX0004)
贵州省科技支撑计划(黔科合支撑[2018]2154)
重庆交通大学研究生科研创新项目(2021B0001)。
关键词
温度效应
变分模态分解(VMD)
K-L散度(KLD)
桥梁挠度分离
健康监测
temperature effect
variational modal decomposition(VMD)
Kullback-Leibler divergence(KLD)
bridge deflection separation
health monitoring
