【摘要】 将二值Logistic回归模型应用到砂土液化预测中,以国内外173组样本数据作为研究对象,通过相关性分析选择震级M、有效应力σ′0、平均粒径D50、修正的CPT锥尖阻力qc1和地震剪应力比SSR5个指标作为敏感性分析指标,对砂土液化指标的敏感性进行了分析。结果表明:二值Logistic回归模型预测准确率达到了90.8%,其概率表达式可用于砂土液化预测;各指标的敏感性排序为M>qc1>σ′0>D50≈SRR;σ′0和qc1指标与砂土液化的相对危险性呈负相关趋势,以qc1指标敏感性较大,M、D50、SRR指标与砂土液化的相对危险性呈正相关趋势,以M敏感性最大。更多还原

【Abstract】 This paper adopts the binary logistic regression model in the prediction of soil liquefaction,selects173 groups of sample data from home and abroad as the research object to establish the binary logistic regression model,and chooses five indicators including earthquake magnitude M,the effective vertical stressσ′0,the mean grain size D50,the measured CPT tip resistance qc and the cyclic stress ratio SSR as the basic indicators for the prediction of soil liquefaction according to correlation analysis.Analysis results show that the binary logistic regression model has a good forecast accuracy of 90.8%,and its expression of probability of soil liquefaction can be used for the prediction of soil liquefaction.The ranking result of the sensitivity of indicators is M>qc1>σ′0>D50≈SRR.The indicators includingσ′0,qc1 are negatively correlated with the relative risk of soil liquefaction while M,D50,SRRare positively correlated with the relative risk of soil liquefaction.更多还原