Support vector machines‐based modelling of seismic liquefaction potential

This paper investigates the potential of support vector machines (SVM)‐based classification approach to assess the liquefaction potential from actual standard penetration test (SPT) and cone penetration test (CPT) field data. SVMs are based on statistical learning theory and found to work well in comparison to neural networks in several other applications. Both CPT and SPT field data sets is used with SVMs for predicting the occurrence and non‐occurrence of liquefaction based on different input parameter combination. With SPT and CPT test data sets, highest accuracy of 96 and 97%, respectively, was achieved with SVMs. This suggests that SVMs can effectively be used to model the complex relationship between different soil parameter and the liquefaction potential. Several other combinations of input variable were used to assess the influence of different input parameters on liquefaction potential. Proposed approach suggest that neither normalized cone resistance value with CPT data nor the calculation of standardized SPT value is required with SPT data. Further, SVMs required few user‐defined parameters and provide better performance in comparison to neural network approach. Copyright © 2006 John Wiley & Sons, Ltd.     

基于地面运动强度及标准贯入试验的上海地区砂土地震液化评价

上海市地处长江三角洲前缘,黄浦江和苏州河交汇区域,特殊的地理环境与沉积环境导致浅部砂层广泛发育。随着城市建设的不断推进,上海城市区域范围的砂土地震液化风险评价成为亟待研究的课题。文章基于上海市工程钻孔数据,结合地震地面运动加速度分布与标准贯入试验,建立区域性地震液化危险性评价模型,对上海市进行了地震液化危险性评价。研究认为当发生50年超越概率10%的地震条件下,上海市陆域面积的66.0%将不会产生地震砂土液化灾害,21.8%的陆域面积仅发生轻微液化,只有崇明、横沙、长兴三岛,黄浦江及苏州河两岸地震液化等级达到中等甚至严重,占全市陆域面积12.3%;50年超越概率2%的地震条件下,随着峰值地面运动加速度整体升高,全市范围内轻微—严重液化区域明显增多,可能发生地震液化的总面积达到全市陆域面积46.25%。上海市存在砂土地震液化的危险性,但是发生概率较低。研究认为,目前的抗震设计规范中上海市的设防烈度偏高,可能导致不必要的建设成本。同时研究中的不同超越概率下的地震液化危险性评价结果为上海市工程建设相关标准的合理化改进的提供了建议和参考。     

Estimating spatial correlation structures based on CPT data

Spatial correlation structures are usually used while modelling inherent uncertainty associated with soil properties. To retrieve this information a sufficient number of measurements and equal intervals between measurements or samples are always required. However, in practice, it is not usual to conduct site investigation in that way subject to a limited budget. Therefore, it is of interest to understand the possible bias created by using samples with unequal intervals. This research focuses on estimating the vertical spatial correlation structures of cone penetration tests (CPT) data conducted at a liquefaction site in Taiwan. The appropriate fitting range used in calibrating the correlation model is also discussed. The results show that cone tip resistance has larger correlation distances than that of sleeve friction. The estimated correlation distances of the CPT profiles are similar while using an equal sampling interval within a certain fitting range (denoted as the critical interval of measurements in this research). The results indicate that the effect of using different curve-fitting ranges is relatively insignificant for correlation distances of sleeve friction, whereas the correlation distances of cone tip resistance are more sensitive to different curve-fitting ranges. Moreover, the mean values of correlation distances are more constant for the equal sampling interval cases than for the unequal interval cases.     

A back analysis scheme for refined soil stratification based on integrating borehole and CPT data

Utilizing both borehole and Cone Penetration Testing (CPT) data in soil stratification helps to get more convincing soil stratification results. However, the soil classification results revealed by borehole (Unified Soil Classification System, USCS) and CPT tests (soil behavior type, SBT) are commonly not consistent. This study proposes a feasible solution to integrate the borehole and CPT data with the tree-based method. The tree-based method is naturally suitable for soil stratification tasks as it aims to divide the subsurface space into several clusters based on the similarities of the soil types. A novel boundary dictionary method is proposed to enhance the model performance on complex soil layer conditions. A probabilistic mapping matrix between the USCS-SBT system is built based on a collected municipal database with collocated borehole and CPT data. The optimal soil stratification results can be selected based on considering multiple borehole information and pruning the structure of trees. The structure of the trees can be optimized in a back analysis perspective with the Sequential Model-Based Global Optimization (SMBO) algorithm which aims to maximize the possibility of observing the borehole information based on the USCS-SBT probabilistic mapping matrix. The uncertainties of the optimal soil stratification results can be estimated based on a weighted Gini index method. The performance of the proposed method is validated based on a real case in New Zealand with a cross-validation method. The results indicate that the proposed method is robust and effective.     

A Bayesian network approach for predicting seismic liquefaction based on interpretive structural modeling

The Bayesian network (BN) is a type of graphical network based on probabilistic inference that has been gradually applied to assessment of seismic liquefaction potential. However, how to construct a robust BN remains underexplored in this field. This paper aims to present an efficient hybrid approach combining domain knowledge and data to construct a BN that facilitates the integration of multiple factors and the quantification of uncertainties within a network model to assess seismic liquefaction. Initially, only using given domain knowledge, a naive network model can be constructed using interpretive structural modeling. Thereafter, some effective information about the naive model is provided to construct a robust model using structural learning of BN from historic data. Finally, the returning predictive results and the predictive results are compared to other methods including non-probabilistic and probabilistic models for seismic liquefaction using the metrics of the overall accuracy, the area under the curve of receiver operating characteristic, prediction, recall and F₁ score. The methodology proposed in this paper achieved better performance, and we discussed the power and value of the proposed approach at the end of this paper, which suggest that BN is a good alternative tool for seismic liquefaction prediction.     

A genetic algorithm approach for assessing soil liquefaction potential based on reliability method

Deterministic approaches are unable to account for the variations in soil’s strength properties, earthquake loads, as well as source of errors in evaluations of liquefaction potential in sandy soils which make them questionable against other reliability concepts. Furthermore, deterministic approaches are incapable of precisely relating the probability of liquefaction and the factor of safety (FS). Therefore, the use of probabilistic approaches and especially, reliability analysis is considered since a complementary solution is needed to reach better engineering decisions. In this study, Advanced First-Order Second-Moment (AFOSM) technique associated with genetic algorithm (GA) and its corresponding sophisticated optimization techniques have been used to calculate the reliability index and the probability of liquefaction. The use of GA provides a reliable mechanism suitable for computer programming and fast convergence. A new relation is developed here, by which the liquefaction potential can be directly calculated based on the estimated probability of liquefaction (P _L ), cyclic stress ratio (CSR) and normalized standard penetration test (SPT) blow counts while containing a mean error of less than 10% from the observational data. The validity of the proposed concept is examined through comparison of the results obtained by the new relation and those predicted by other investigators. A further advantage of the proposed relation is that it relates P _L and FS and hence it provides possibility of decision making based on the liquefaction risk and the use of deterministic approaches. This could be beneficial to geotechnical engineers who use the common methods of FS for evaluation of liquefaction. As an application, the city of Babolsar which is located on the southern coasts of Caspian Sea is investigated for liquefaction potential. The investigation is based primarily on in situ tests in which the results of SPT are analysed.     

隐伏断层在强震砂土液化中的作用——以2008年汶川Mw 7.9地震为例

2008年汶川Mw7.9地震的强地面震动在龙门山前地区造成大量的砂土液化、喷砂冒水等地震灾害现象。震后野外调查发现,砂土液化点主要分布于地下水位只有几米深的山前河流的低阶地处,以大面积砾性土液化为特征,约58%的液化点位于距北川断层20~35km的范围内。对喷水高度及喷水过程进行了详细记录,喷水高度与峰值加速度并没有明显的相关性,喷水高度异常点(2m)集中于山前断裂系统近地表投影处。汶川地震中喷水高度异常、砾性土液化的位置与山前断裂系统的吻合性说明,沉积盆地内的地质构造可能在砂土液化强度和与震动相关的地震灾害方面起到促进作用,所以在类似的地质和水文环境中,除主震的断层错动外,应考虑地质构造在地震危险性评估和建筑物抗震设计中的重要作用。     

Evaluation of liquefaction potential based on CPT results using evolutionary polynomial regression

In this paper a new approach is presented, based on evolutionary polynomial regression (EPR), for determination of liquefaction potential of sands. EPR models are developed and validated using a database of 170 liquefaction and non-liquefaction field case histories for sandy soils based on CPT results. Three models are presented to relate liquefaction potential to soil geometric and geotechnical parameters as well as earthquake characteristics. It is shown that the EPR model is able to learn, with a very high accuracy, the complex relationship between liquefaction and its contributing factors in the form of a function. The attained function can then be used to generalize the learning to predict liquefaction potential for new cases not used in the construction of the model. The results of the developed EPR models are compared with a conventional model as well as a number of neural network-based models. It is shown that the proposed EPR model provides more accurate results than the conventional model and the accuracy of the EPR results is better than or at least comparable to that of the neural network-based models proposed in the literature. The advantages of the proposed EPR model over the conventional and neural network-based models are highlighted.     

汶川地震中甘肃清水田川黄土液化的试验研究

2008年5·12特大地震中,位于甘肃省清水县郭川乡的田川村发生了饱和黄土的液化滑移灾害。本文首先在对田川场地进行考察的基础上,综合田川在汶川地震中的震害情况以及滑移区的地形条件,将该地区在汶川地震中的烈度进行了修正。其次对田川黄土进行了物性指标测试以及室内动三轴液化试验,根据试验结果,综合考虑产生液化所需的场地及土性条件、黄土的动强度和液化特性,对田川黄土液化灾害进行了分析,并采用反应分析的方法对其进行了液化判定。研究结果证明了田川黄土液化的事实存在性,为低烈度区黄土液化提供了新的震害依据。

     

基于人工神经网络的砂土液化势评价

本文利用静力触探（ＣＰＴ）场地液化数据,建立了液化势判定的反向传播神经网络模型,研究表明,同传统方法相比,人工神经网络方法在判别砂土液化势方面是可行的。     

基于CFD的地震液化研究新进展

综述了近年来关于液化土体流体动力学特征的试验发展状况,以及基于计算流体动力学（简称CFD）的地震液化数值模拟现状,重点介绍了目前比较活跃的可以较高精度模拟液化土体流动状态的三次伪质点数值方法（简称CIP法）。通过对CFD和传统固体力学在地震液化研究中的应用比较,指出了应用CFD的三大优势,即土体大变形问题、液化土体参数分析以及液化土体中结构物的变形应用CFD分析,均可获得较好的结果。进一步提出,在地震液化应用中,未来CFD的发展应该考虑整合液化前的土体性状研究和地震液化中桩-土-结构物的综合分析。     

基于静力触探试验的液化判别新方法

近来地震液化灾害频发,再次成为研究重点,发展具有良好应用前景的基于静力触探试验（CPT）的液化判别方法对预防液化灾害具有重要意义。以Boulanger数据库171组数据为回归样本,分析既有方法存在的问题,提出了基于CPT液化判别的双曲线模型和计算公式,并通过提取2011年新西兰地震147组液化新数据,对该方法进行对比检验。研究表明,我国岩土工程勘察规范的CPT液化判别方法对浅埋砂层偏于保守,对深层土又明显偏于危险,而国际上具有代表性的Robertson方法,其液化临界线存在低烈度区不合理回弯、高烈度区又偏于保守的问题。提出的新公式在不同地震动强度和砂层埋深下均可给出合理判别结果,克服了国内外既有方法的缺点,并纳入到具有样板规范性质的《建筑工程抗震性态设计通则》修订稿中,可为我国相关规范修订和工程应用提供支持。     

基于贝叶斯网络的地震液化概率预测分析

基于解释结构模型和因果图法,选取12个具有代表性的定性和定量因素,在大量数据不完备的情况下提出了建立贝叶斯网络液化模型的方法。以2011年日本东北地区太平洋近海地震液化不完备数据为例,采用总体精度、ROC曲线下面积、准确率、召回率和F1值5项指标对模型进行综合评估,并与径向基神经网络模型进行对比。结果表明：贝叶斯网络液化模型的回判和预测效果都优于径向基神经网络模型,且对于数据缺失的样本的预测效果也较理想。此外,该模型对于不同土质的液化评估均有较好的适用性。分类不均衡和抽样偏差会对模型的学习和预测效果产生很大影响,建议应同时采用上述5项评估指标进行综合评估模型的优劣。     

液化与非液化场地加速度反应谱对比

通过对2011年新西兰发生的6.3级地震中获取的硬土场地、软土场地和液化场地强震记录实测资料进行分析,研究包括3种类型场地的地震动特征及其相互关系。分析的强震记录为震中距小于50 km,且峰值大于0.05g的23个场地上的加速度记录,其中含3个硬场地、11个软场地和9个液化场地。3种类型场地上的放大系数谱对比分析表明,硬土场地、软土场地和液化场地上的地震动特征具有明显区别,三者层次清晰,差别显著,从地震动表现上液化场地已构成一个与常规场地并列的独立单元;液化场地减少地震动高频分量但同时对低频分量显著放大,与非液化场地相比,液化场地可使其上短周期结构反应减小一半,但同时可使其上长周期结构反应放大2.5～5.0倍;土层的液化对此次地震中克莱斯特彻奇市中心CTV大楼的破坏应有很大影响,大楼自振周期约为0.7 s,地震中土层液化使场地加速度反应谱卓越周期由0.1～0.3 s增到0.5～1.0 s,与大楼的自振周期趋于吻合,加重了大楼震害。以此为鉴,按现有规范中地震动的设计方法,如遇液化场地将对长周期结构给出明显危险的结果,因此从振动角度今后进行结构抗震设计时,可液化场地上的地震动应给予特殊考虑。     

天津泰达地区粉土振动液化特性的研究

本文以动三轴试验和原位测试数据为基础，以天津地区砂性土为例，探讨砂性土振动液化机理及孔隙水压力变化规律，采用多种方法分析判别砂性土液化，为高烈度地区重要工程建筑抗震设计提供重要数据。     

双桥静力触探法判别上海薄夹层粘土地基液化研究

通过对11项典型工程场地进行原位取土及双桥静力触探原位测试分析,重点研究了上海地区薄层粘性土（或粘质粉土）夹层对液化判别的影响,统计分析了锥尖阻力qc、摩阻比Rf与土层粘粒含量的相关关系,提出了完全依据双桥静力触探试验的地基液化判别方法,在工程应用中取得了显著经济效益。     

Modification of the liquefaction potential index and liquefaction susceptibility mapping for a liquefaction-prone area (Inegol,Turkey)

Although some liquefaction assessment methods were proposed to evaluate the liquefaction potential of sandy soils, the conventional method based on the standard penetration test (SPT) has been commonly used in most countries and in Turkey. However, it alone is not a sufficient tool for the evaluation of liquefaction potential. The liquefaction potential index was proposed to quantify the severity of liquefaction. Nevertheless, the liquefaction potential index and the severity categories do not answer the question: "Which areas will not liquify?" Besides, the categories do not include a "moderate" category; on the other hand, the "high" and "low" categories are included. This situation is also contrary to the nature of classification schemes. In this study, the liquefaction potential index and the liquefaction potential categories were modified by considering the existing form of the categories based on the liquefaction potential index. While the category of low was omitted, the categories of moderate and "non-liquefied" were adopted. A factor of safety of 1.2 was assumed as the lowest value for the liquefaction potential category of non-liquefied. In addition, the town of Inegol in the Marmara region became the case study for checking the performance of the liquefaction potential categories suggested in this study.     

基于逐步判别分析的砂土液化预测研究

为高效地进行砂土液化的预测,运用逐步判别法,从8个液化影响因子中选择平均粒径、烈度、震中距等3个判别能力显著的影响因子,建立判别函数,并利用工程实例进行验证。研究结果表明：逐步判别分析模型预测性能良好,且能有效地选择对砂土液化起主导作用的因子。相比距离判别分析,逐步判别分析建立的判别函数更加稳定,且所需测试因子较少,节省了因试验和现场调查所耗费的大量人力、物力和时间,因此逐步判别分析是一种值得推广的砂土液化预测方法。