基于奇异值分解信号消噪技术.doc

基于奇异值分解的信号消噪技术 摘 要 模态参数识别是从结构不同位置的动力响应信号中提取出结构的模态参数，即：从动力测试响应信号数据中确定结构的模态参数(模态振型、固有频率和阻尼比)。每一个结构都有其固有的模态参数，并且如果结构动力特性发生变化了，那么结构的模态参数也将发生相应的变化。显见，结构的模态参数识别是非常重要的，为诊断结构健康状况提供了依据。 基于输出的模态参数识别方法利用的信息主要是系统的自由振动信号，要获 得自由振动信号首先需获得结构的响应信号。由于环境激励的不充分和噪声等干 扰因素的存在，导致信号测试信号不能直接用于参数辨识，需要对信号进行消噪处理。即从大量背景噪声中提取出可用于模态参数辨识的有用信号成分,剔除干扰因素，提取有用信息。此时，信号消噪技术研究变得尤为重要。 本文采用了一种将Hankel矩阵和奇异值分解相结合的消噪方法。该方法首先对测量信号构造的Hankel矩阵进行奇异值分解，再利用测量信号快速傅立叶变换结果中主频率的个数来确定有效秩阶次，接着通过消噪信号的信噪比和均方差大小确定重构矩阵结构，最后通过反对角线平均法得到消噪后的信号数据。通过数值仿真，对不同信号进行定秩和消噪，从结果可以知道这种方法具有较好的消噪效果。 关键词：信号消噪；奇异值分解；快速傅立叶变换；信噪比；均方差 A Method for Noise Reduction Based on Singular Value Decomposition Abstract Accurate estimate of the modal parameters of an offshore structure is crucial to many practical engineering issues, such as finite element (FE) model updating and validation, damage detection, etc. Modal parameter identification method uses the the response signal of structure ,but actual response signal often contains a lot of noise, which will affect the accuracy of signal recognition. The test signal de-noising processing is an important step in signal processing. Using Singular Value Decomposition（SVD）of constructed Hankel matrix by measured signal is an effective method for eliminating the random noise. The key is to choose the rank of the Hankel matrix and determine the structure of the reconstruction matrix. In this paper, it is using the number of the main frequency in the result of using signal fast Fourier transform to determine the rank of the Hankel matrix, and through HYPERLINK app:ds:SNRSNR(HYPERLINK app:ds:SignalSignal HYPERLINK app:ds:toto HYPERLINK app:ds:NoiseNoise HYPERLINK app:ds:RatioRatio) and MSE(HYPERLINK app:ds:meanMean HYPERLINK app:ds:squareSquare HYPERLINK app:ds:errorError) to determine reconstruction matrix structure. Simulation and experiment validated this method. The results shows that the number of rank is double of the main frequency, and the best lines of reconstruction matrix is half of the len