Development of numerical simulation method for prediction of surface earthquake fault behavior

本研究的目的是在模型实验的基础上，建立一种新的断层行为预测的数值分析方法，以阐明断层破裂过程的机理。开发了CT扫描和三维光弹性实验新技术；三维光弹性采用了新的器件和数据分析方法。模型实验表明，断裂的控制机制是导致断裂行为高度变化的分岔。基于这一观察，一种新的分析方法被称为非线性随机有限元法（NL-SSFEM）。该方法分析了松散土层的随机行为，找到了与实际断层相对应的最不稳定分岔解。NL-SSFEM在应用土工材料方面进行了调整。通过复制其他研究人员进行的几个模型实验，验证了该方法的有效性。对于正断层或逆断层的二维问题和横向滑动的三维问题，NL-SSFEM能够较好地评价断层形态和载荷的若干参数。应该强调的是，这些参数的变异性也得到了很好的评估，并且NL-SSFEM能够计算出失效的概率。下一步是对实际断层的研究。靶区为野岛断裂（反向和横向耦合断裂）和四龙岗断裂（反向断裂）。NL-SSFEM建立了实际松散层的计算机模型，对断层进行了模拟，并将模拟断层的形态与观测断层进行了比较。良好的一致性支持了NL-SSFEM作为预测故障行为工具的有效性。NL-SSFEM提供了有关断层概率的信息，如导致地表断层的基底移动量。虽然没有经过验证，但从过去的断层数据来看，概率值似乎是合理的。

This research is aimed at developing a new numerical analysis method for prediction of fault behavior, based on model experiments which are to clarify the mechanism of rupture processes of faulting. New experiment techniques using CT scanner and 3D photoelasticity are developed ; the 3D photoelasticity uses new devices and data analysis method. It is shown in the model experiments that the governing mechanism of faulting is the bifurcation that leads to highly variable fault behaviors. Based on this observation, a new analysis method, called non-linear stochastic finite element method (NL-SSFEM) has been developed. This method analyzes the stochastic behavior of unconsolidated soil layers and find the most unstable bifurcated solution that corresponds to the actual faulting.The NL-SSFEM is tuned in applying geo-materials. The validity of the method is verified by reproducing several model experiments carried out by other researchers. For two-dimensional problem of normal or reverse faulting and for three-dimensional problem of lateral sliding, the NL-SSFEM is able to give fairly good evaluation of several parameters of fault configuration and loading. It should be emphasized the variability of these parameters is also well evaluated and that the NL-SSFEM is able to compute the probability of failure. Actual faulting is next studied. The targets are the Nojima Fault (coupled reverse and lateral faulting) and the Shirlongpong Fault (reverse faulting). Constructing a computer model for actual unconsolidated layers, the NL-SSFEM simulates the faulting, and the configuration of simulated faults is compared with observed ones. Good agreement supports the usefulness of the NL-SSFEM as a tool for predicting the fault behavior. The NL-SSFEM provides information about the faulting probability, as the amount of base movement that leads to the surface faulting. Although not verified, the values of the probability seems reasonable from the past data of faulting.