Conventional predictive deconvolution assumes that the reflection coefficients of the earth conform to an uncorrelated white noise sequence. The Wiener-Hopf (WH) equation is constructed to solve the filter and elimina...Conventional predictive deconvolution assumes that the reflection coefficients of the earth conform to an uncorrelated white noise sequence. The Wiener-Hopf (WH) equation is constructed to solve the filter and eliminate the correlated components of the seismic records, attenuate multiples, and improve seismic resolution. However, in practice, the primary refl ectivity series of fi eld data rarely satisfy the white noise sequence assumption, with the result that the correlated components of the primary reflectivity series are also eliminated by traditional deconvolution. This results in signal distortion. To solve this problem, we have proposed an improved method for deconvolution. First, we estimated the wavelet correlation from seismic records using the spectrum-modeling method. Second, this wavelet autocorrelation was used to construct a new autocorrelation function which contains the correlated components caused by the existence of multiples and avoids the correlated components of the primary reflectivity series. Finally, the new autocorrelation function was brought into the WH equation, and the predictive fi lter operator was calculated for deconvolution. In this paper, we have applied this new method to simulated and field data processing, and we have compared its performance with that of traditional predictive deconvolution. Our results show that the new method can adapt to non-white refl ectivity series without changing the statistical characteristics of the primary reflection coefficient series. Compared with traditional predictive deconvolution, the new method reduces processing noise and improves fidelity, all while maintaining the ability to attenuate multiples and enhance seismic resolution.展开更多
Empirical mode decomposition (EMD) is proposed to identify linear structure under non-stationary excitation,and non-white noise coefficient is introduced under the assumption of random signals consisting of white nois...Empirical mode decomposition (EMD) is proposed to identify linear structure under non-stationary excitation,and non-white noise coefficient is introduced under the assumption of random signals consisting of white noise and non-white noise signals. The cross-correlation function of response signal is decomposed into mode functions and residue by EMD method. The identification technique of the modal parameters of single freedom degree is applied to each mode function to obtain natural frequencies, damping ratios and mode shapes. The results of identification of the five-degree freedom linear system demonstrate that the proposed method is effective in identifying the parameters of linear structures under non-stationary ambient excitation.展开更多
为实现基于振动传递比函数的工作模态分析方法能够在任一荷载工况下识别结构模态参数,引入参考响应思路,构建响应功率谱传递比(Power Spectral Density Transmissibility, PSDT)函数。首先利用比例函数的极限定理,揭示PSDT在系统极点处...为实现基于振动传递比函数的工作模态分析方法能够在任一荷载工况下识别结构模态参数,引入参考响应思路,构建响应功率谱传递比(Power Spectral Density Transmissibility, PSDT)函数。首先利用比例函数的极限定理,揭示PSDT在系统极点处的重要特性,进而根据这一特性建立PSDT驱动的峰值法;同时为解决传统传递比方法无法识别结构阻尼的问题,建立基于PSDT驱动的最小二乘复频域法(LSCF),通过参数化拟合思路识别频率、振型和阻尼比,并运用稳定图辅助剔除虚假模态。通过10层剪切型框架结构数值算例,对比研究外部激励性质对PSDT法及传统频域法(峰值法、频域分解法)识别结果的影响。最后,运用PSDT法对环境激励下的人行桥进行工作模态分析,并与传统响应传递比方法及随机子空间法(SSI)结果进行对比。研究结果表明:在同一工况下不同参考响应的PSDT函数在系统极点与外部激励性质无关,且等价于振型比值;PSDT法相比于传统频域法对外部激励具有更为良好的鲁棒性,能够降低识别谐波激励引起的虚假模态的风险;不同于传统响应传递比方法,在任一工况下基于PSDT法能够识别人行桥的包括阻尼比在内的工作模态参数,并产生更为清晰的峰值和稳定图,具有更好的可操作性;该方法识别结果与SSI结果吻合较好,验证了其在任一荷载工况下分析实际桥梁结构工作模态特性的可行性。展开更多
基金supported by Scientific Research Foundation of Shandong University of Science and Technology for Recruited Talents(No.2017RCJJ034)
文摘Conventional predictive deconvolution assumes that the reflection coefficients of the earth conform to an uncorrelated white noise sequence. The Wiener-Hopf (WH) equation is constructed to solve the filter and eliminate the correlated components of the seismic records, attenuate multiples, and improve seismic resolution. However, in practice, the primary refl ectivity series of fi eld data rarely satisfy the white noise sequence assumption, with the result that the correlated components of the primary reflectivity series are also eliminated by traditional deconvolution. This results in signal distortion. To solve this problem, we have proposed an improved method for deconvolution. First, we estimated the wavelet correlation from seismic records using the spectrum-modeling method. Second, this wavelet autocorrelation was used to construct a new autocorrelation function which contains the correlated components caused by the existence of multiples and avoids the correlated components of the primary reflectivity series. Finally, the new autocorrelation function was brought into the WH equation, and the predictive fi lter operator was calculated for deconvolution. In this paper, we have applied this new method to simulated and field data processing, and we have compared its performance with that of traditional predictive deconvolution. Our results show that the new method can adapt to non-white refl ectivity series without changing the statistical characteristics of the primary reflection coefficient series. Compared with traditional predictive deconvolution, the new method reduces processing noise and improves fidelity, all while maintaining the ability to attenuate multiples and enhance seismic resolution.
基金National Natural Science Foundation(No.19972016)for partly supporting this work
文摘Empirical mode decomposition (EMD) is proposed to identify linear structure under non-stationary excitation,and non-white noise coefficient is introduced under the assumption of random signals consisting of white noise and non-white noise signals. The cross-correlation function of response signal is decomposed into mode functions and residue by EMD method. The identification technique of the modal parameters of single freedom degree is applied to each mode function to obtain natural frequencies, damping ratios and mode shapes. The results of identification of the five-degree freedom linear system demonstrate that the proposed method is effective in identifying the parameters of linear structures under non-stationary ambient excitation.
文摘为实现基于振动传递比函数的工作模态分析方法能够在任一荷载工况下识别结构模态参数,引入参考响应思路,构建响应功率谱传递比(Power Spectral Density Transmissibility, PSDT)函数。首先利用比例函数的极限定理,揭示PSDT在系统极点处的重要特性,进而根据这一特性建立PSDT驱动的峰值法;同时为解决传统传递比方法无法识别结构阻尼的问题,建立基于PSDT驱动的最小二乘复频域法(LSCF),通过参数化拟合思路识别频率、振型和阻尼比,并运用稳定图辅助剔除虚假模态。通过10层剪切型框架结构数值算例,对比研究外部激励性质对PSDT法及传统频域法(峰值法、频域分解法)识别结果的影响。最后,运用PSDT法对环境激励下的人行桥进行工作模态分析,并与传统响应传递比方法及随机子空间法(SSI)结果进行对比。研究结果表明:在同一工况下不同参考响应的PSDT函数在系统极点与外部激励性质无关,且等价于振型比值;PSDT法相比于传统频域法对外部激励具有更为良好的鲁棒性,能够降低识别谐波激励引起的虚假模态的风险;不同于传统响应传递比方法,在任一工况下基于PSDT法能够识别人行桥的包括阻尼比在内的工作模态参数,并产生更为清晰的峰值和稳定图,具有更好的可操作性;该方法识别结果与SSI结果吻合较好,验证了其在任一荷载工况下分析实际桥梁结构工作模态特性的可行性。
文摘基于协方差驱动随机子空间识别(covariance-driven stochastic subspace identification,SSI-COV)方法的模态参数识别具有强鲁棒性、高精度的优势,在结构工作模态分析中应用广泛。为保证模态参数识别的准确性,新近提出的基于随机子空间(stochastic subspace identification,SSI)的模态参数不确定性量化方法,可有效估计模态参数的方差,但由于其计算各中间变量时,需显式表示出Jacobian矩阵,导致矩阵运算维度高、计算效率低。为此,提出一种基于SSI-COV的模态参数不确定度高效计算方法。首先,计算振动响应相关函数的方差,通过奇异值分解(singular value decomposition,SVD),选取合适的奇异值截断阶数,由每阶奇异向量组装出多组Hankel矩阵的扰动。其次,根据一阶矩阵摄动理论,隐式计算SSI-COV算法各中间变量的一阶扰动,最终,由多组模态参数的扰动叠加计算出方差。最后,以桁架结构模型为例,采用所提方法辨识结构模态参数并计算模态参数方差,分析了Hankel矩阵维度及相关函数奇异值截断阶数对辨识结果的影响,结果表明计算得到的模态参数方差与蒙特卡洛仿真(Monte Carlo simulation,MCS)结果非常接近,且模态参数不确定度可作为剔除虚假模态的有效依据。
