基于L1正则化逻辑回归模型的速度大脉冲预测及影响因素分析

Prediction and influencing factors of big velocity pulse based on L1 regularized logistic regression model

包含速度大脉冲的地震动与普通地震动有显著的不同,此类地震动对建筑结构有着特殊的破坏作用。有效预测速度大脉冲是否发生和分析影响速度大脉冲发生的因素,对于概率地震危险性分析和减轻地震灾害有重要的作用。首先从美国NGA数据库中选取315条强震动记录,经过预处理后得到研究所需的289条记录,基于处理后的强震动记录并结合相对频度分析方法研究不同因素对速度大脉冲的影响。然后,利用L1正则化逻辑回归方法建立速度大脉冲预测模型,模型最优评价指标的接受者操作特征曲线下方面积为0.76;并对模型进行影响因素敏感性分析,发现其对于破裂区的距离最为敏感。最后,选取符合模型数据分布规律的35条汶川地震实测数据对建立的预测模型进行验证,其中30条地震记录预测正确,相较于已有模型,预测的准确率有一定提高。结果表明,文章建立的速度大脉冲预测模型有较好的准确性和可靠性。

Ground motions containing big velocity pulse,which are significantly different from ordinary ground motions,have a special destructive effect on building structures.Scientific prediction of the occurrence of big velocity pulse and analysis of factors affecting the occurrence of big velocity pulse play an important role in probabilistic seismic risk analysis and reduction of earthquake disasters.In this paper,315 strong motion records were selected from the NGA database of the United States,and 289 records were obtained after preprocessing.Based on the processed strong motion records,the effects of different factors on big velocity pulse were studied with the relative frequency analysis method.Then the L1 regularized logistic regression method was used to establish a prediction model of big velocity pulse,and the AUC(area under curve)value of the model is 0.76.It was found that the model was the most sensitive to the distance of fracture zone through a sensitivity analysis of the influencing factors.Finally,the prediction model was verified by 35 items of Wenchuan earthquake data,and the results indicated that 30 earthquake records were predicted correctly.The results showed that compared with the existing models,the big velocity pulse prediction model proposed in this paper has better accuracy and reliability.