摘要
针对光纤陀螺在面板堆石坝面板挠度监测中易受到噪声干扰,难以准确提取反映结构变形特征信号的实际问题,提出一种基于最小二乘平滑滤波与CEEMDAN混合降噪的方法。该方法运用CEEMDAN将光纤陀螺实测信号进行分解,得到一系列IMF分量。分别对每一阶IMF分量进行傅里叶频谱分析得到幅值谱图和幅值的方差,根据幅值方差的大小判断噪声IMF分量与有用信号IMF分量的分界,结合最小二乘平滑滤波对噪声IMF分量进行降噪。最后将降噪后的IMF分量与有用信号IMF分量进行重构,得到降噪后的光纤陀螺信号。通过对仿真信号和水布垭面板堆石坝面板挠度监测的实测数据进行分析,该方法能有效滤除噪声信号,准确提取反映结构变形的特征信号,验证了该方法对实际工程中光纤陀螺测量信号降噪的可行性和适用性。
In view of the fact that fiber optic gyroscope is easy to be disturbed by noise in the monitoring of concrete face deflection of concrete face rockfill dam,and it is difficult to accurately extract the characteristic signal of structural deformation, a noise reduction method based on least square smooth filtering and CEEMDAN was proposed.In this method, the measured signal of fiber optic gyroscope was decomposed by CEEMDAN, and a series of IMF components were obtained.The fourier spectrum analysis of each order IMF component was carried out to get the amplitude spectrum and the variance of the amplitude. According to the magnitude of the amplitude variance, the boundary between the noise IMF component and the useful signal IMF component was judged, and the noise IMF component was reduced by combining the least square smooth filtering.Finally, the denoised IMF component and the useful signal IMF component were reconstructed to get the denoised fiber optic gyroscope signal.Through the analysis of the simulation signal and the measured data of the concrete face deflection monitoring of Shuibuya concrete face rockfill dam, this method can effectively filter the noise signal and accurately extract the characteristic signal reflecting the structural deformation,the feasibility and applicability of this method for noise reduction of fiber optic gyroscope measurement signal in practical engineering were verified.
作者
徐朗
蔡德所
XU Lang;CAI Desuo(College of Hydraulic and Environmental Engineering,China Three Gorges University,Yichang 443002,China)
出处
《振动与冲击》
EI
CSCD
北大核心
2020年第10期269-278,共10页
Journal of Vibration and Shock
基金
国家自然科学基金(59879002)。
关键词
面板挠度监测
集合经验模态分解
傅里叶频谱分析
方差
最小二乘平滑滤波
deflection monitoring of concrete face
complete ensemble empirical mode decomposition with adaptive noise
fourier spectrum analysis
variance
least square smooth filtering
