基于机器学习-网格搜索优化的砂土液化预测

摘要
图/表
参考文献(36)
相关文章 (13)

全文: PDF (4612 KB) HTML (1 KB)
输出: BibTeX | EndNote (RIS)

摘要砂土液化是一种破坏力较强的地震次生灾害，传统的砂土液化判别方法存在一定的局限性。提出两种液化判别方法，第一种是基于新西兰岩土数据库（New Zealand Geotechnical Database，NZGD）中519组静力触探试验数据，建立具有砂土液化预测功能的机器学习模型。首先建立支持向量机（Support Vector Machine，SVM）、随机森林（Random Forest，RF）、XGboost（eXtreme Gradient Boosting，XGB）三种机器学习分类模型，运用网格搜索法（GridSearchCV）进行超参数优化后，采用整体精度（OA）、精确率（P）、召回率（R）、F1值对模型进行性能评估，对历史液化数据进行模型验证并与国内外方法进行结果对比。第二种是基于历史震害数据，采用经验判断法确定的静力触探初判条件。结果表明：随机森林可作为一种具有较强的预测能力的液化判别模型，通过与国内《岩土工程勘察规范》以及国际Olsen方法进行对比，选取要素简便且计算迅速的随机森林能够达到与上述两种权威方法接近的准确性，是一种可实行的液化判别模型；此外，基于历史液化数据库归结出不同烈度下，具备液化埋深限制的锥尖阻力阈值，经数据验证在7烈度区、8烈度区、9烈度区的准确率良好，与《岩土工程勘察规范》进行对比发现有可操作性好、可解释性强、可适用性广等优点。所建立的模型对砂土液化预测具有较强的适用性，静力触探初判条件亦可作为快速液化判别的参考值，二种方法结合可对科学研究和工程建设提供较好的参考价值。

	服务

	把本文推荐给朋友
	加入我的书架
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	王昭栋1
	王自法2
	3
	李兆焱2
	苗鹏宇1
	吴禄源1

关键词 ：砂土液化, 机器学习, 液化预测, 静力触探初判条件, 网格搜索

Abstract：Two liquefaction discrimination methods are proposed. The first method is to establish a machine learning model with sand liquefaction prediction function based on 519 sets of static penetration test data from the New Zealand Geotechnical Database (NZGD). First, three machine learning classification models, namely, Support Vector Machine (SVM), Random forest (RF), and XGboost (eXtreme Gradient Boosting (XGB), are established. After super parameter optimization using GridSearchCV, the performance of the model is evaluated using the overall accuracy (OA), accuracy (P), recall (R), and F1 values, Conduct model validation on historical liquefaction data and compare the results with domestic and foreign methods. The second method is based on historical earthquake damage data and uses empirical judgment to determine the initial judgment conditions for static penetration testing. The results show that Random forest can be used as a liquefaction Discriminative model with strong prediction ability. By comparing with the domestic Geotechnical investigation and the international Olsen method, Random forest with simple elements and rapid calculation can achieve the accuracy close to the above two power methods, which is a feasible liquefaction Discriminative model; In addition, based on the historical liquefaction database, the conic resistance threshold method with the limitation of liquefaction burial depth under different intensities is summarized. The accuracy rate of the method is good in the 7 intensity zone, 8 intensity zone, and 9 intensity zone through data verification. Compared with the Code for Geotechnical investigation, it has the advantages of good operability, strong interpretability, and wide applicability. The established model has strong applicability for predicting soil liquefaction, and the initial judgment conditions of static cone penetration can also serve as reference values for rapid liquefaction discrimination. The combination of the two methods can provide good reference value for scientific research and engineering construction.

Key words： Soil liquefaction Machine learning Liquefaction prediction Initial judgment conditions for static penetration testing GridSearchCV;

收稿日期: 2023-03-08 出版日期: 2024-03-15

引用本文:

王昭栋1,王自法2,3,李兆焱2,苗鹏宇1,吴禄源1. 基于机器学习-网格搜索优化的砂土液化预测 [J]. 振动与冲击, 2024, 43(5): 82-93.
WANG Zhaodong1,WANG Zifa2,3,LI Zhaoyan2,MIAO Pengyu1,WU Luyuan1. Prediction of sandy soil liquefaction based on machine learning-GridSearchCV. JOURNAL OF VIBRATION AND SHOCK, 2024, 43(5): 82-93.

链接本文:

http://jvs.sjtu.edu.cn/CN/ 或 http://jvs.sjtu.edu.cn/CN/Y2024/V43/I5/82

[1]张思宇,李兆焱,袁晓铭.基于静力触探试验的液化判别新方法[J].岩土力学,2022,43(06):1596-1606.DOI:10.16285/j.rsm.2021.1524. ZHANG Si-yu, LI Zhao-yan, YUAN Xiao-ming, . A new method for evaluating liquefaction based on cone penetration test[J]. Rock and Soil Mechanics, 2022, 43(6): 1596-1606.( in Chinese) [2]Robertson P K. Evaluation of flow liquefaction and liquefied strength using the cone penetration test: an update[J]. Canadian Geotechnical Journal, 2022, 59(4): 620-624 [3]王长虹,汤道飞,王昆等.静力触探试验的宏细观耦合分析方法与应用[J].岩土力学,2021,42(07):1815-1827.DOI:10.16285/j.rsm.2020.1724. WANG Chang-hong, TANG Dao-fei, WANG Kun, WU Zhao-xin. Macro and micro coupling analysis method and application of cone penetration test[J]. Rock and Soil Mechanics, 2021, 42(7): 1815-1827.( in Chinese) [4]舒坦,续海金,章军锋等.北京房山地区太平山褶皱的变形特征和形成时代:华北克拉通早白垩世挤压构造的意义[J].地球科学,2019,44(05):1734-1748. Shu Tan, Xu Haijin, Zhang Junfeng, Liu Qiang, 2019. Deformation Characteristics and Time of Taipingshan Folds in Fangshan Area, Beijing: Implications for Early Cretaceous Compressional Tectonics of North China Craton. Earth Science, 44(5): 1734-1748. doi: 10.3799/dqkx.2018.119.( in Chinese) [5]Gou, Qiyang, et al. "The effect of tectonic deformation and preservation condition on the shale pore structure using adsorption-based textural quantification and 3D image observation." Energy 219 (2021): 119579. .DOI:https://doi.org/10.1016/j.energy.2020.119579. [6]Eberhart‐Phillips, Donna, et al. "The Influence of Basement Terranes on Tectonic Deformation: Joint Earthquake Travel‐Time and Ambient Noise Tomography of the Southern South Island, New Zealand." Tectonics 41.4 (2022): e2021TC007006.DOI:https://doi.org/10.1029/2021TC007006. [7]闫志晓,李雨润,张健.饱和砂土中直群桩及土体地震动力响应特征研究[J].振动与冲击,2020,39(18):44-53.DOI:10.13465/j.cnki.jvs.2020.18.006. YAN Zhixiao，LI Yurun，ZHANG Jian. Seismic response characteristics of straight group piles and soil in saturated sand. JOURNAL OF VIBRATION AND SHOCK, 2020, 39(18): 44-53. ( in Chinese) [8]王建宁,付继赛,庄海洋,窦远明,马国伟.可液化场地中复杂异跨地铁地下车站结构的地震反应分析[J].振动与冲击,2020,39(07):170-179.DOI:10.13465/j.cnki.jvs.2020.07.024 WANG Jianning, FU Jisai, ZHUANG Haiyang, DOU Yuanming1, MA Guowei1. Seismic response analysis of complex subway station structure with unequal-span in liquefiable foundation. JOURNAL OF VIBRATION AND SHOCK, 2020, 39(7): 170-179.( in Chinese) [9]赵凯,王秋哲,王彦臻,庄海洋,陈国兴. 可液化地基地下结构地震反应特征简化有效应力分析[J]. 振动与冲击, 2021, 40(21): 39-46. ZHAO Kai, WANG Qiuzhe, WANG Yanzhen, ZHUANG Haiyang, CHEN Guoxing. Effects of soil-underground structure interaction on seismic response of liquefiable sit around underground structure. JOURNAL OF VIBRATION AND SHOCK, 2021, 40(21): 39-46. ( in Chinese) [10]王自法, 廖吉安, 王延伟, 位栋梁, 赵登科. 2023. 基于深层卷积神经网络的震级快速估算方法. 地球物理学报, 66(1): 272-288, doi: 10.6038/cjg2022P0709 OF VIBRATION AND SHOCK, 2023, 42(8): 177-185. WANG ZiFa, LIAO JiAn, WANG YanWei, WEI DongLiang, ZHAO DengKe. 2023. A fast magnitude estimation method based on deep convolutional neural networks. Chinese Journal of Geophysics (in Chinese), 66(1): 272-288, doi: 10.6038/cjg2022P0709 [11]Wu L, Wang Z, Ma D, et al. A continuous damage statistical constitutive model for sandstone and mudstone based on triaxial compression tests[J]. Rock Mechanics and Rock Engineering, 2022, 55(8): 4963-4978. [12]Wu L, Ma D, Wang Z, et al. An deep CNN-based constitutive model for describing of statics characteristics of rock materials[J]. Engineering Fracture Mechanics, 2023: 109054. [13]李程程,李兆焱,袁晓铭.区域土壤地震液化预测简化方法[J].地球物理学报,2020,63(05):2084-2095. LI ChengCheng, LI ZhaoYan, YUAN XiaoMing. 2020. Simplified prediction method for regional seismic soil liquefaction. Chinese Journal of Geophysics (in Chinese), 63(5): 2084-2095, .DOI:10.6038/cjg2020N0052. [14]姜礼涛,周爱红,袁颖,刘育林,宁志杰,牛建广.基于NRS-ISSA-SVM的砂土液化判别模型[J].地震工程学报,2022,44(03):570-578.DOI:10.20000/j.1000-0844.20220118001. JIANG Litao, ZHOU Aihong, YUAN Ying, LIU Yulin, NING Zhijie, NIU Jianguang. A discriminant model for sand liquefaction based on NRS-ISSA-SVM[J]. China Earthquake Engineering Journal,2022,44(3):570-578.( in Chinese) [15]Kohestani, et al. "Evaluation of liquefaction potential based on CPT data using random forest." NAT HAZARDS (2015). [16]Vapnik V N. An overview of statistical learning theory[J]. IEEE transactions on neural networks, 1999, 10(5): 988-999.DOI: 10.1109/72.788640 [17]Breiman. Random forests[J]. MACH LEARN, 2001, 2001,45(1)(-):5-32. [18]Chen T , Guestrin C . XGBoost: A Scalable Tree Boosting System[J]. ACM, 2016.DOI: https://doi.org/10.1145/2939672.2939785 [19]Moss R E S. CPT-based probabilistic assessment of seismic soil liquefaction initiation[M]. University of California, Berkeley, 2003. [20]Olsen R S . Cyclic liquefaction based on the cone penetrometer test[J]. National Center for Earthquake Engineering Research, 1997. [21]中华人民共和国建设部. GB 50021－2001（2009年版）岩土工程勘察规范[S]. 北京: 中国建筑工业出版社,2009. Ministry of Construction of the People’s Republic of China. GB 50021 － 2001(2009 version) Code for investigation of geotechnical engineering[S]. Beijing:China Architecture and Building Press, 2009.( in Chinese) [22]Hooker S, Erhan D, Kindermans P J, et al. Evaluating feature importance estimates[J]. 2018. [23]Altmann A, Toloşi L, Sander O, et al. Permutation importance: a corrected feature importance measure[J]. Bioinformatics, 2010, 26(10): 1340-1347 [24]Liam M. Wotherspoon and Michael J. Pender and Rolando P. Orense. Relationship between observed liquefaction at Kaiapoi following the 2010 Darfield earthquake and former channels of the Waimakariri River[J]. Engineering Geology, 2012.DOI:https://doi.org/10.1016/j.enggeo.2011.11.001 [25]Beyzaei, Christine, Z, et al. Laboratory-based characterization of shallow silty soils in southwest Christchurch[J]. Soil Dynamics & Earthquake Engineering, 2018.DOI:https://doi.org/10.1016/j.soildyn.2018.01.046 [26]Bastin S H, Quigley M C, Bassett K. Paleoliquefaction in Christchurch, New Zealand[J]. Bulletin, 2015, 127(9-10): 1348-1365. [27]Villamor P, Almond P, Tuttle M P, et al. Liquefaction features produced by the 2010–2011 Canterbury earthquake sequence in southwest Christchurch, New Zealand, and preliminary assessment of paleoliquefaction features[J]. Bulletin of the Seismological Society of America, 2016, 106(4): 1747-1771. [28]Heritage R , Kupec J . LIQUEFACTION-RESISTANT FOUNDATIONS FOR RESIDENTIAL BUILDINGS[C]// International Young Geotechnical Engineering Conference. 2013.DOI:10.3233/978-1-61499-297-4-497 [29]Lees J J, Ballagh R H, Orense R P, et al. CPT-based analysis of liquefaction and re-liquefaction following the Canterbury earthquake sequence[J]. Soil Dynamics and Earthquake Engineering, 2015, 79: 304-314. [30]Beyzaei C Z, Bray J D, Cubrinovski M, et al. Laboratory-based characterization of shallow silty soils in southwest Christchurch[J]. Soil Dynamics and Earthquake Engineering, 2018, 110: 93-109. [31]Green R A, Maurer B W, Cubrinovski M, et al. Assessment of the relative predictive capabilities of CPT-based liquefaction evaluation procedures: Lessons learned from the 2010-2011 Canterbury earthquake sequence[C]//Proc. 6th Intern. Conf. on Earthquake Geotechnical Engineering. 2015. [32]Geyin M, Maurer B W, Bradley B A, et al. CPT-based liquefaction case histories compiled from three earthquakes in Canterbury, New Zealand[J]. Earthquake Spectra, 2021, 37(4): 2920-2945. [33]Cox S C , Ballegooy S V , Rutter H K , et al. Can artesian groundwater and earthquake-induced aquifer leakage exacerbate the manifestation of liquefaction?[J]. Engineering Geology, 2020, 281:105982.DOI:https://doi.org/10.1016/j.enggeo.2020.105982 [34]GB/T 17742-2020, 中国地震烈度表[S]. GB/T 17742-2020, China Seismic Intensity Scale [S].( in Chinese) [35]GB 50011-2010, 建筑抗震设计规范[S]. GB 50011-2010，Code for seismic design of buildings[S]( in Chinese) [36]李兆焱,袁晓铭,曹振中等.基于新疆巴楚地震调查的砂土液化判别新公式[J].岩土工程学报,2012,34(03):483-489. LI Zhao-yan, YUAN Xiao-ming, CAO Zheng-zhong, SUN Rui, DONG Lin, SHI Jiang-hua. New evaluation formula for sand liquefaction based on survey of Bachu Earthquake in Xinjiang. Chinese J. Geot. Eng., 2012, 34(3): 483-437.( in Chinese)