Classification evaluation of construction sites with thick overburden based on machine learning
-
摘要:
针对因测量等误差对等效剪切波速计算的影响而造成的场地类别容易因单个因素稍有变化即发生的类别改变问题,从江苏省盐城地区收集了大量厚覆盖土层情况下的标准贯入值、深度、剪切波速等相关现场试验数据,利用机器学习方法进行训练建模,研究多特征值模型解决厚覆盖土层情况下场地分类问题的能力。结果表明:随机森林模型的分类精度在加入“等效变异系数”后可达97.7%,且其泛化能力以及对样本总体的判断能力均优于支持向量机模型,该模型为厚覆盖土层建筑场地类别的判断提供了一种新的方式。将二次判断结果与勘探报告结果对比,结果证明该随机森林模型可用于场地分类变化问题的二次判断,为避免工程现场在类似情况下出现过于保守的判断提供了可靠的依据。
Abstract:In response to the problem that the category of the site is easily changed due to slight changes in a single factor caused by measurement and other errors in the calculation of equivalent shear wave velocity, a large amount of relevant field test data such as standard penetration value, depth, and shear wave velocity were collected under thick overburdens of Yancheng area of Jiangsu Province. Machine learning methods were used for training and modeling, and the ability of multi eigenvalue models to solve site classification problems under thick overburdens was studied. The results showed that through feasibility analysis, the accuracies of the logistic regression model, the support vector machine model and the random forest model were 0.809, 0.939, 0.951, respectively. Considering the accuracy gap between each two models of the above three, the support vector machine algorithm and the random forest algorithm were selected as the optimal algorithms for building the model. In order to consider the integrity of the entire borehole as much as possible, this paper proposes a parameter called as “equivalent coefficient of variation”, which effectively improves the accuracy of the model. Subsequently, when establishing the support vector machine model, the classification performance of linear, polynomial, and Gaussian kernels was compared, and the Gaussian kernel function was ultimately selected for model building. The accuracy of the obtained support vector machine model was 0.951. When establishing a random forest model, the classification performance of the model was tested by setting different numbers of decision trees. Finally, 150 decision trees were selected to build the model, and the accuracy of the obtained random forest model was 0.977. From the results, the accuracy of the support vector machine model and the random forest model are 95.1% and 97.7%, respectively, with recall rates of 98.2% and 97.3%. The AUC (area under curve) values of both models are 0.98. Therefore, while the classification performance of the random forest model is not inferior to that of the support vector machine model, it has a higher adaptability to the sample data, and the recall and accuracy of the random forest model are similar, that is, the model’s judgment on the sample population is more balanced. In summary, the above random forest model is optimal to solve the problem studied in this paper, and can provide reliable basis for determining the category of sites with thick overburdens. Therefore the random forest model was used to determine the site category of 75 sets of data in the critical sample of this study. The results showed that 61 sets were consistent with the judgment results of the exploration report, while 14 sets were different from the judgment results of the exploration report. Moreover, the model’s judgment on class Ⅲ sites was completely consistent with the exploration report. All above proves that the model not only has excellent judgment ability in non-critical situations, but also maintains good judgment ability when used to solve problems in critical situations. Therefore, this model can make secondary judgments for similar engineering problems and provide effective reference basis. Based on the random forest model, the judgment results are output in sequence, and are organized and verified according to the original drilling information. It is found that in the judgment on the critical sample, the model correctly judged eight drilling holes, and only two drilling holes had different site classification judgments from the exploration report, all of which were classified as class Ⅳ drilling sites in the report and were classified as class Ⅲ drilling sites in the model. In practical engineering, judgments made on site for safety reasons are often conservative, and such judgments are magnified as two different site classification results near the boundary. This can explain the significant divergence between the model and exploration report’s judgments on class Ⅳ sites.
-
-
表 1 不同核函数的模型精度
Table 1 Accuracy of the models with different kernel functions
核函数 精度 线性核 83.6% 多项式核 (d=2) 91.2% 多项式核 (d=3) 93.7% 高斯核 95.1% 表 2 随机森林模型输出结果与现场结果对比
Table 2 Comparison between the output results of random forest model and on-site results
钻孔序号 现场判断结果 模型判断结果 1 Ⅳ类 Ⅲ类 2 Ⅳ类 Ⅲ类 3 Ⅳ类 Ⅳ类 4 Ⅲ类 Ⅲ类 5 Ⅲ类 Ⅲ类 6 Ⅲ类 Ⅲ类 7 Ⅲ类 Ⅲ类 8 Ⅲ类 Ⅲ类 9 Ⅲ类 Ⅲ类 10 Ⅲ类 Ⅲ类 -
陈国兴,丁杰发,方怡,彭艳菊,李小军. 2020. 场地类别分类方案研究[J]. 岩土力学,41(11):3509–3522. Chen G X,Ding J F,Fang Y,Peng Y J,Li X J. 2020. Investigation of seismic site classification scheme[J]. Rock and Soil Mechanics,41(11):3509–3522 (in Chinese).
陈卓识,袁晓铭,孙锐,王克. 2019. 土层剪切波速不确定性对场地刚性判断的影响[J]. 岩土力学,40(7):2748–2754. Chen Z S,Yuan X M,Sun R,Wang K. 2019. Impact of uncertainty in in-situ shear-wave velocity on the judgement of site stiffness[J]. Rock and Soil Mechanics,40(7):2748–2754 (in Chinese).
迟明杰,李小军,陈学良,马笙杰. 2021. 场地划分中存在的问题及建议[J]. 地震学报,43(6):787–803. Chi M J,Li X J,Chen X L,Ma S J. 2021. Problems and suggestions on site classification[J]. Acta Seismologica Sinica,43(6):787–803 (in Chinese).
黄鑫怀,李增华,邓腾,刘志锋,陈冠群,曾皓轩,郭世超. 2023. 基于机器学习的华南诸广山花岗岩体铀矿潜力评价[J]. 地球科学,48(12):4427–4440. Huang X H,Li Z H,Deng T,Liu Z F,Chen G Q,Zeng H X,Guo S C. 2023. Uranium potential evaluation of Zhuguangshan granitic pluton in South China based on machine learning[J]. Earth Science,48(12):4427–4440 (in Chinese).
黄衍,查伟雄. 2012. 随机森林与支持向量机分类性能比较[J]. 软件,33(6):107–110. Huang Y,Zha W X. 2012. Comparison on classification performance between random forests and support vector machine[J]. Software,33(6):107–110 (in Chinese).
姬建,王乐沛,廖文旺,张卫杰,朱德胜,高玉峰. 2021. 基于WUS概率密度权重法的边坡稳定系统可靠度分析[J]. 岩土工程学报,43(8):1492–1501. Ji J,Wang L P,Liao W W,Zhang W J,Zhu D S,Gao Y F. 2021. System reliability analysis of slopes based on weighted uniform simulation method[J]. Chinese Journal of Geotechnical Engineering,43(8):1492–1501 (in Chinese).
赖成光,陈晓宏,赵仕威,王兆礼,吴旭树. 2015. 基于随机森林的洪灾风险评价模型及其应用[J]. 水利学报,46(1):58–66. Lai C G,Chen X H,Zhao S W,Wang Z L,Wu X S. 2015. A flood risk assessment model based on random forest and its application[J]. Journal of Hydraulic Engineering,46(1):58–66 (in Chinese).
林凤仙,段继平,许峻,李正光,许昭永. 2020. 判定场地土类别的等效剪切波速度的最佳计算深度[J]. 工程地球物理学报,17(2):166–176. Lin F X,Duan J P,Xu J,Li Z G,Xu Z Y. 2020. The optimum calculation depth for determination of the site soil classification by equivalent shear wave velocity[J]. Chinese Journal of Engineering Geophysics,17(2):166–176 (in Chinese).
刘方园,王水花,张煜东. 2018. 支持向量机模型与应用综述[J]. 计算机系统应用,27(4):1–9. Liu F Y,Wang S H,Zhang Y D. 2018. Overview on models and applications of support vector machine[J]. Computer Systems &Applications,27(4):1–9 (in Chinese).
刘益平,邓维祥,周康. 2022. 基于标贯试验的岩土剪切波速多因素公式分析[J]. 中国勘察设计,(增刊):30–33. Liu Y P,Deng W X,Zhou K. 2022. Analysis of multi factor formula for shear wave velocity of rock and soil based on standard penetration test[J]. China Engineering Consulting,(S2):30−33 (in Chinese).
罗路广,裴向军,崔圣华,黄润秋,朱凌,何智浩. 2021. 九寨沟地震滑坡易发性评价因子组合选取研究[J]. 岩石力学与工程学报,40(11):2306–2319. Luo L G,Pei X J,Cui S H,Huang R Q,Zhu L,He Z H. 2021. Combined selection of susceptibility assessment factors for Jiuzhaigou earthquake-induced landslides[J]. Chinese Journal of Rock Mechanics and Engineering,40(11):2306–2319 (in Chinese).
王昊,严加永,付光明,王栩. 2020. 深度学习在地球物理中的应用现状与前景[J]. 地球物理学进展,35(2):642–655. doi: 10.6038/pg2020CC0476 Wang H,Yan J Y,Fu G M,Wang X. 2020. Current status and application prospect of deep learning in geophysics[J]. Progress in Geophysics,35(2):642–655 (in Chinese).
许冲,徐锡伟. 2012. 逻辑回归模型在玉树地震滑坡危险性评价中的应用与检验[J]. 工程地质学报,20(3):326–333. Xu C,Xu X W. 2012. Logistic regression model and its validation for hazard mapping of landslides triggered by Yushu earthquake[J]. Journal of Engineering Geology,20(3):326–333 (in Chinese).
战吉艳,陈国兴,刘建达. 2012. 苏州城区场地等效剪切波速计算深度取值探讨[J]. 地震工程与工程振动,32(5):166–171. Zhan J Y,Chen G X,Liu J D. 2012. Discussion on calculation depth selection of equivalent shear wave velocity for site classification in urban area of Suzhou[J]. Earthquake Engineering and Engineering Vibration,32(5):166–171 (in Chinese).
张学工. 2000. 关于统计学习理论与支持向量机[J]. 自动化学报,26(1):32–42. Zhang X G. 2000. Introduction to statistical learning theory and support vector machines[J]. Acta Automatica Sinica,26(1):32–42 (in Chinese).
中华人民共和国住房和城乡建设部.2010. 建筑抗震设计规范(GB50011—2010) [M]. 北京:中国建筑工业出版社:13. Ministry of Housing and Urban Rural Development of the People’s Republic of China. 2010. Code for Seismic Design of Buildings (GB50011−2010)[M]. Beijing:China Construction Industry Press:13 (in Chinese).
Altmann A,Toloşi L,Sander O,Lengauer T. 2010. Permutation importance:A corrected feature importance measure[J]. Bioinformatics,26(10):1340–1347. doi: 10.1093/bioinformatics/btq134
Bajaj K,Anbazhagan P. 2019. Seismic site classification and correlation between VS and SPT-N for deep soil sites in Indo-Gangetic basin[J]. J Appl Geophys,163:55–72. doi: 10.1016/j.jappgeo.2019.02.011
Bhavsar H,Panchal M H. 2012. A review on support vector machine for data classification[J]. Int J Adv Res Comput Eng Technol,1(10):185–189.
Breiman L. 2001. Random Forests[J]. Machine Learning, 45 :5−32.
Calderón‐Macías C,Sen M K,Stoffa P L. 2000. Artificial neural networks for parameter estimation in geophysics[J]. Geophys Prospect,48(1):21–47. doi: 10.1046/j.1365-2478.2000.00171.x
Chelgani S C,Matin S S,Hower J C. 2016. Explaining relationships between coke quality index and coal properties by Random Forest method[J]. Fuel,182:754–760. doi: 10.1016/j.fuel.2016.06.034
Ching J. 2020. Value of geotechnical big data and its application in site-specific soil property estimation[J]. J GeoEng,15(4):173–182.
Liu H J,Wang Y N,Lu X F. 2005. A method to choose kernel function and its parameters for support vector machines[C]//Proceedings of 2005 International Conference on Machine Learning and Cybernetics Vol. 7. Guangzhou:IEEE:4277−4280.
Martens D,De Backer M,Haesen R,Vanthienen J,Snoeck M,Baesens B. 2007. Classification with ant colony optimization[J]. IEEE Trans Evol Computat,11(5):651–665. doi: 10.1109/TEVC.2006.890229
Phoon K K,Zhang W G. 2023. Future of machine learning in geotechnics[J]. Georisk:Assess Manage Risk Eng Syst Geohazards,17(1):7–22.
Tesfamariam S,Liu Z. 2010. Earthquake induced damage classification for reinforced concrete buildings[J]. Struct Saf,32(2):154–164. doi: 10.1016/j.strusafe.2009.10.002
Xiao S H,Zhang J,Ye J M,Zheng J G. 2021. Establishing region-specific N-VS relationships through hierarchical Bayesian modeling[J]. Eng Geol,287:106105. doi: 10.1016/j.enggeo.2021.106105
Zhang J,Wang T P,Xiao S H,Gao L. 2021a. Chinese code methods for liquefaction potential assessment based on standard penetration test:An extension[J]. Soil Dyn Earthq Eng,144:106697. doi: 10.1016/j.soildyn.2021.106697
Zhang R H,Li Y Q,Goh A T C,Zhang W G,Chen Z X. 2021b. Analysis of ground surface settlement in anisotropic clays using extreme gradient boosting and random forest regression models[J]. J Rock Mech Geotech Eng,13(6):1478–1484. doi: 10.1016/j.jrmge.2021.08.001
Zhang W G,Phoon K K. 2022. Editorial for advances and applications of deep learning and soft computing in geotechnical underground engineering[J]. J Rock Mech Geotech Eng,14(3):671–673. doi: 10.1016/j.jrmge.2022.01.001