基于随机森林算法的砂土液化预测方法

doi:10.11720/wtyht.2020.1501

摘要
图/表
参考文献
相关文章
Metrics

全文: PDF(721 KB) HTML
输出: BibTeX | EndNote (RIS)

摘要

砂土液化的影响因素较多且复杂。以唐山大地震的72个场地的实测液化样本数据为例,在不丢失任何信息的前提下,选取了8个砂土液化的判别指标,通过计算样本数据的Gini系数,采用CART算法的决策树对数据的特征属性进行划分。在此基础之上,通过增加多个决策树构造随机森林的方式,在一定程度上降低了单个决策树学习过度造成的过拟合风险,同时,通过10轮交叉验证的方式确定了决策树的最大高度为5,随机森林中决策树的个数为20时,模型的效果达到最佳。研究结果表明,与抗震设计规范中的标贯试验法判别公式相比,决策树模型和随机森林模型的训练结果和预测结果有显著提高,尤其是随机森林模型在训练样本和预测样本上均没有出现误判,稳定性更高。

	服务

	把本文推荐给朋友
	加入引用管理器
	E-mail Alert
	RSS
	作者相关文章
	彭刘亚
	解惠婷
	冯伟栋

关键词 ：砂土液化, 判别指标, 决策树, 随机森林

Abstract：

Among a variety of complicated factors that are related to sand liquefaction,8 discriminant factors have been picked out of 72 samples in the earthquake event happened in Tangshan without losing any tiny but useful information.By calculating Gini coefficient with CART algorithm,a decision tree has been undertaken to divide the features of original sample dataset.Moreover,in order to reduce overfitting risk of a single decision tree,random forest with multiple trees have been created.Meanwhile,with 10-fold cross validation,best estimators with 5 max-depth and 20 trees can perform with much more stable and reliable results.The research shows that,compared to standard penetration test from Code for seismic design of buildings,both decision tree and random forest have a better predicting precision, especially there have been no false classifications with higher stability using random forest model.

Key words： sand liquefaction discriminant indicator decision tree random forest

收稿日期: 2019-11-25 出版日期: 2020-12-29

P631.4

基金资助:中国地震局三结合课题(3JH202002013)

作者简介: 彭刘亚(1990-),男,工程师,主要从事工程地震及地震灾害现场调查等方面的研究工作

引用本文:

彭刘亚, 解惠婷, 冯伟栋. 基于随机森林算法的砂土液化预测方法[J]. 物探与化探, 2020, 44(6): 1429-1434.
PENG Liu-Ya, XIE Hui-Ting, FENG Wei-Dong. The method of predict sand liquefaction based on random forest algorithm. Geophysical and Geochemical Exploration, 2020, 44(6): 1429-1434.

链接本文:

https://www.wutanyuhuatan.com/CN/10.11720/wtyht.2020.1501 或 https://www.wutanyuhuatan.com/CN/Y2020/V44/I6/1429

Fig.1 决策树分类模型示意

Fig.2 随机森林分类模型示意

序号	判别指标								液化情况
序号	I	L/km	D₅₀/mm	C_u	d_w/m	d_s/m	N_63.5/击	τ_d/ $σ V'$	液化情况
1	7	68.6	0.410	2.90	1.09	4.15	5	0.1000	1
2	7	83.3	0.187	4.00	1.20	2.45	8	0.0900	1
3	7	83.3	0.111	2.02	0.80	1.35	6	0.0800	1
?	?	?	?	?	?	?	?	?	?
19	7	79.0	0.120	1.55	1.37	3.60	19	0.0940	0
20	7	81.2	0.160	2.67	1.05	4.30	12	0.1050	0
?	?	?	?	?	?	?	?	?	?
26	8	116.4	0.200	2.70	1.60	8.70	8	0.2120	1
27	8	116.4	0.170	1.91	3.30	5.80	5	0.1600	1
?	?	?	?	?	?	?	?	?	?
43	8	70.9	0.300	2.43	2.30	12.30	13	0.2030	0
44	8	47.0	0.310	2.42	2.00	3.46	8	0.1630	0
45	8	117.0	0.073	7.50	1.53	11.90	26	0.2170	0
?	?	?	?	?	?	?	?	?	?
50	9	22.0	0.200	1.94	0.43	2.61	10	0.4620	1
51	9	22.0	0.240	2.08	1.15	4.50	22.2	0.4150	1
?	?	?	?	?	?	?	?	?	?
62	9	14.0	0.160	2.25	4.90	9.38	61	0.3180	0
63	9	9.6	0.210	3.15	3.50	8.35	31	0.3470	0
64	9	11.0	0.160	2.76	4.50	4.50	22	0.2480	0

Table 1 砂土液化训练样本集

序号	判别指标								液化情况
序号	I	L/km	D₅₀/mm	C_u	d_w/m	d_s/m	N_63.5/击	τ_d/ $σ V'$	液化情况
1	7	76.8	0.166	1.65	0.50	1.70	3	0.1000	1
2	7	60.8	0.360	3.30	1.59	6.65	23	0.1030	0
3	7	70.0	0.145	8.50	0.85	1.80	2	0.0890	1
4	7	49.0	0.140	2.31	1.00	4.80	14	0.1080	0
5	7	81.2	0.140	1.60	1.40	4.35	9	0.1000	1
6	8	116.0	0.265	2.81	3.30	13.80	17	0.1900	0
7	8	117.4	0.134	2.23	3.20	7.20	8	0.1720	1
8	9	17.0	0.185	1.90	0.61	3.80	4	0.4580	1

Table 2 砂土液化测试样本集

Fig.3 决策树分类过程示意

Fig.4 交叉验证示意

Fig.5 模型训练(a)及预测结果(b)

[1]	高大钊, 袁聚云. 土质学与土力学[M]. 北京: 人民交通出版社, 2001.
[1]	Gao D Z, Yuan J Y. Soil science and soil mechanics[M]. Beijing: China Communications Press, 2001.
[2]	宫凤强, 李嘉维. 基于PCA-DDA原理的砂土液化预测模型及应用[J]. 岩土力学, 2016,37(s1):448-453.
[2]	Gong F Q, Li J W. Discrimination model of sandy soil liquefaction based on PCA-DDA principle and its application[J]. Rock and Soil Mechanics, 2016,37(s1):448-453.
[3]	刘章军, 叶燎原, 彭刚. 砂土地震液化的模糊概率评判方法[J]. 岩土力学, 2008,29(4):876-880.
[3]	Liu Z J, Ye L Y, Peng G. Fuzzy probability comprehensive evaluation method for sand liquefaction during earthquake[J]. Rock and Soil Mechanics, 2008,29(4):876-880.
[4]	谢君斐. 关于修改抗震规范砂土液化判别式的几点意见[J]. 地震工程与工程振动, 1984,4(2):95-109.
[4]	Xie J F. Some opinions on the modification of sand liquefaction discriminant of seismic code[J]. Earthquake Engineering and Engineering Dynamics, 1984,4(2):95-109.
[5]	陈国兴, 孔梦云, 李小军, 等. 以标贯试验为依据的砂土液化确定性及概率判别法[J]. 岩土力学, 2015,36(1):9-26.
[5]	Chen G X, Kong M Y, Li X J, et al. Deterministic and probabilistic triggering correlations for assessment of seismic soil liquefaction at nuclear power plant[J]. Rock and Soil Mechanics, 2015,36(1):9-26.
[6]	刘红军, 薛新华. 砂土地震液化预测的人工神经网络模型[J]. 岩土力学, 2004,25(12):1942-1946.
[6]	Liu H J, Xue X H. Artificial neural network model for prediction of seismic liquefaction of sand soil[J]. Rock and Soil Mechanics, 2004,25(12):1942-1946.
[7]	刘勇健. 基于聚类—二叉树支持向量机的砂土液化预测模型[J]. 岩土力学, 2008,29(10):2764-2768.
[7]	Liu Y J. Support vector machine model for predicting sand liquefaction based on clustering binary tree algorithm[J]. Rock and Soil Mechanics, 2008,29(10):2764-2768.
[8]	张菊连, 沈明荣. 基于逐步判别分析的砂土液化预测研究[J]. 岩土力学, 2010,31(s1):298-301.
[8]	Zhang J L, Shen M R. Sand liquefaction prediction based on stepwise discriminant analysis[J]. Rock and Soil Mechanics, 2010,31(s1):298-301.
[9]	刘年平, 王宏图, 袁志刚, 等. 砂土液化预测的Fisher判别模型及应用[J]. 岩土力学, 2012,33(2):554-557.
[9]	Liu N P, Wang H T, Yuan Z G, et al. Fisher discriminant analysis model of sand liquefaction and its application[J]. Rock and Soil Mechanics, 2012,33(2):554-557.
[10]	温博文, 董文瀚, 解武杰, 等. 基于改进网格搜索算法的随机森林参数优化[J]. 计算机工程与应用, 2018,54(10):154-157.
[10]	Wen B W, Dong W H, Xie W J, et al. Parameter optimization method for random forest based on improved grid search algorithm[J]. Computer Engineering and Applications, 2018,54(10):154-157.
[11]	张亮, 宁芊. CART决策树的两种改进及应用[J]. 计算机工程与设计, 2015,36(5):1209-1214.
[11]	Zhang L, Ning Q. Two improvements on CART decision tree and its application[J]. Computer Engineering and Design, 2015,36(5):1209-1214.

[1]	郭建宏, 张占松, 张超谟, 周雪晴, 肖航, 秦瑞宝, 余杰. 用地球物理测井资料预测煤层气含量——基于斜率关联度—随机森林方法的工作案例[J]. 物探与化探, 2021, 45(1): 18-28.
[2]	刘崇民, 马生明, 朱立新, 梁胜跃, 李冰. 甘肃小铁山隐伏多金属矿床地球化学预测指标[J]. 物探与化探, 2009, 33(6): 646-651.

Viewed

Full text

Abstract

Cited

Shared

Discussed