Abstract
Simulation analysis is an effective method for the safety assessment of a building after an earthquake. This kind of analysis is effective for evaluating foundations or piles, whose safety cannot be confirmed by visual check.
In conducting these analyses, an input motion should be set to reproduce an observed wave at an observed point of the analysis model. If the analysis model is in a linear range, the input motion can easily be set using the frequency domain method, which is the most popular method. However, this setting is difficult if the analysis model (e.g., soil) exhibits non-linear behavior. In these cases, the properties of the analysis model changes at every time step. Therefore, the frequency domain method cannot be applied. Instead, the time domain method must be applied, but no effective inversion method of the time domain is available.
This study proposes a new time domain inversion method. First, the method outline and details are described. Second, the method precision is confirmed through experimentation. Thus, a new method for an input motion inversion in a non-linear soil model is presented.
Followings are a summary of the study findings:
(1) This study proposes a new time domain inversion method with a generalized inverse of the perturbation impulse matrix and the iterative method. In proposed method, first, effects of perturbation impulse inputs are calculated. Secondly, “perturbation impulse matrix” is made from these effect vectors. Then, from singular value decomposition of perturbation impulse matrix, general inverse of the perturbation impulse matrix composed by main modes is made. Using this inverse, input correction amount is calculated. And then these corrections repeated until error become small enough.
(2) Through a sample problem, the accuracy of this proposed method is confirmed. Then, it is confirmed that an identified input motion reproduce a target output motion in surface of elasto-plastic soil model.
The findings of this study confirm the applicability of this proposed method for inversion. There is, however, the need for further improvement of the procedure because it is difficult to select effective modes. Moreover, much calculation time that this method need is important an important issue.
