Time Domain Structural Damage Detection Based On Multichannel Singular Spectrum Analysis

The structural damage detection becomes an important topic in civil engineering,because the safety of the structure is directly related to human lives,property and social development.Time domain structural damage detection method is easy to measure dynamic signals,can provide more detection equations,has no intermediate process error and can be directly applied to the nonlinear dynamic system,which makes it more applicable.But the traditional time domain structural damage detection method is inaccuracy with measurement noise effect,and a new time domain structural damage detection method was proposed in this dissertation based on multichannel singular spectrum analysis to reduce the effect of noise.The following aspects are discussed through theoretical analysis,numerical simulation and experimental validation:(1)The distribution of noise information and damage information in multichannel singular spectral components were studied,and it concludes that the lower order components contain the least measurement noise and the most damage information.The indexes of noise information content and damage information content in multichannel singular spectral components were proposed.The finer components could be extracted to detect local damages in the structure,which can reduce the noise effect.(2)Based on the above results,a new time domain structural damage detection method was proposed and analyzed theoretically,in which the correlation of measurement responses in different locations were considered.The identification equation based on the finer components of multichannel singular spectral analysis was proposed based on the subspace projection method and traditional time domain sensitivity method.(3)The factors of the proposed identification method were investigated.The different projection subspaces and selection method of finer components were discussed,and it concludes the multichannel singular spectral analysis components of analytical response construct the finer projection subspace.The window lengths were discussed,and it concludes that the larger finer region is obtained by longer window length.(4)The proposed method was validated by numerical simulation of a two dimensional truss model with 20% and 30% measurement noise,and the damage identification results were compared with those from the traditional sensitivity method.The results show that the proposed method is effective and steady with high level measurement noise.A seven-storey steel frame test was employed to verify the practicability of the proposed method.