A bivariable regression correction method for acceleration integration considering lossless time phases to evaluate the soil cyclic shear behavior in centrifuge tests

Abstract

Using integral displacements from acceleration records, the inversion analysis of soil cyclic shear behavior has been extensively employed in modeling tests and in-situ monitoring of site seismic response. However, there is a lack of attention to the reliability of correction methods for double-integral accelerograms to achieve displacements, especially the phase problem. This might be one of the prime reasons in data processing that lead to non-closure and discontinuity in the hysteresis loops, resulting in significant discreteness in the shear modulus and damping ratio. To address this problem, a novel correction method is proposed using a locally weighted regression function to correct both static parts and dynamic parts in the time domain while integrating acceleration records. The abundant records of paired accelerations and displacements have proved that the phase drift of integral displacements is visibly solved by the proposed method. Meanwhile, the method has high precision in integral amplitudes. The double-integral correction methods have a significant influence on the characteristics and shapes of the hysteresis loops, which is mainly manifested in the phase coordination of shear stress and shear strain. The proposed method can solve the problems of incomplete, discontinuous, and irregular hysteresis loops caused by data processing. The inverse analysis of the shear modulus and damping ratio by the proposed method is less dispersive, which is consistent with the established patterns of what is already known and the results of existing studies. The proposed method is of great theoretical importance and application value to improve the accuracy of the cyclic shear stress-strain response of soils.