地震地质 ›› 1981, Vol. 3 ›› Issue (3): 71-82.

• 科研简报 • 上一篇    

地震时砂土液化的数理统计预测

朱淑莲   

  1. 国家地震局地质研究所
  • 收稿日期:1979-11-01 出版日期:1981-09-09 发布日期:2009-11-25

MATHEMATICO-STATISTICAL PREDICTION OF LIQUEFACTION OF SOIL DURING AN EARTHQUAKE

Zhu Shulian   

  1. Institute of Geology, National Bureau of Seismology
  • Received:1979-11-01 Online:1981-09-09 Published:2009-11-25

摘要: 1976年唐山地震期间,附近一些地区出现了砂土液化现象。本文根据工程勘探中的实例及实验数据,进行了砂土液化作用的统计和综合预测。文章中采用贝叶斯准则下的逐步判别分析方法,对唐山地区已知液化地点和非液化地点,作了五种检验计算,挑选出准确度及可靠性最高的变量模式。从而,对于当地震为Ⅷ度烈度时的砂土液化进行了预测。五种预测结果相同,互相印证,可靠性高,分组最大后验概率大多在0.99以上。在此条件下,外推预测结果可信。工程实践中,迫切需要对砂土液化进行综合性预测,而一般采用地质学或试验方法,仅能对砂土液化作单因素或少量因素下的预测,本文提出了综合性的通用预测方法,为砂土液化统计预测提供了新途径。砂土液化是平原地区的一种重要地震灾害。国内外多次大地震中,都曾因饱水砂土受到地震作用,引起孔隙水压增高及砂粒间的结合力和摩擦力降低,而使砂层发生液化状态的流动,并伴随有地基承载能力降低或失效。1964年以来,新潟地震及阿拉斯加地震时,由砂土液化造成了罕见的灾害。激发了各国、尤其是美国和日本对砂土液化预测的研究。但还多是单因素或少量因素影响下的分析方法。未能进行综合分析及推断。采用数理统计方法,可以综合分析影响砂土液化的各因素的数字特征。本文采用了一种贝叶斯(Beyes)准则下的多组判别分析,通过对已知液化和非液化类型的地质特征的数据分析,统计归纳,找出其规律性。再与未知地点的相应地质条件的数值特征进行对比,从而得出砂土液化与否的推论。算题过程中,曾得到王启鸣、徐道一、高维安等同志的协助,保定冶金勘探公司提供了占探及试验资料,在此一并感谢。

Abstract: The Tangshan 1976' earthquake (M = 7.8) has caused the extensive soil liquefaction in the broad Beijing-Tianjin-Tangshan plain region resulting in a severe damage. For this reason we have to estimate whether the liquefaction would happen during the future shocks that should be an important basis for proposing the measures of reduction of seismic hazard. As is well known, the engineering construction is in urgent need of the comprehensive prediction of liquefaction. Usually, the conventional geological or experimental methods could only give the predictable indications in the cases of simple and less factors. This paper statistically predicts the soil liquefaction on the basis of the measured and tested data during the engineering exploration.Using the method of stepwise discrimination analysis of Beyes’ criterion 13 and 15 variances selected from the liquidated and unliquidated localities can be used as correlation calculations in five ways. The variance pattern with the highest accuracy and confidence is chosen to predict the liquefaction at a location to be a VI11 level in intensity. The five kinds of prediction are in a close agreement and can be proved each other. Moreover, the maximum probability posterior of grouped samples is mostly above 0.99 with high confidence. In such a case, the extrapolated results are reliable.