Determination of geotechnical parameters based on the unsupervised learning method
-
摘要: 随着城市工程建设的发展,建筑工程事故问题愈发突出,采用传统方求取的岩土参数区间无法满足实际工程需要。基于无监督学习思想,选取工程性质较差的泥炭质土,结合工程经验选用8个物理指标作为输入集,利用主成分分析(PCA)算法实现多样本多参数去耦合的降维处理,得出各物理指标相关性及敏感度,结合其相关性及敏感度赋予不同埋深泥炭质土物理指标的综合评价值。利用k-means聚类分析泥炭质土物理指标、综合评价值及工程特性之间关系,为岩土参数选取提供理论基础。采用监督学习方法—BP神经网络算法分析无监督结果,验证(PCA—k-means)算法模型的合理性。将通过聚类分析得到的正态样本利用多种截尾法优化,得到可靠取值区间,并将取值结果与实际工程取值比较,验证该模型工程参数取值合理性。该算法模型具有较好的工程应用价值,所得研究结果可为工程勘察、设计、施工参数取值提供参考,也能为岩土参数取值分析提供新的分析方法。Abstract: Abstrct : With the development of urban engineering construction, the issue of construction engineering accidents has become more and more prominent. The geotechnical parameter interval obtained by using the traditional methods cannot meet the needs of actual engineering. Based on the idea of unsupervised learning, the peaty soil with the worst engineering properties was considered, and 8 physical indexes were selected as the input set. The principal component analysis (PCA) algorithm is used to realize the dimensionality reduction of multi-sample and multi-parameter decoupling, and the correlation and sensitivity of each physical index is obtained. Combined with its correlation and sensitivity, the comprehensive evaluation value of physical indexes of peat soil with different buried depths is given. The k-means clustering is used to analyze the relationship among physical index, and comprehensive evaluation value and engineering characteristics of peaty soil provides a theoretical basis for the selection of geotechnical parameters. The supervised learning method-BP neural network algorithm is used to analyze the unsupervised results and verify the accuracy of the (PCA—k-means) algorithm model. The normal samples obtained by clustering analysis are optimized by a variety of truncation methods to obtain a reliable value range, and the value results are compared with the actual engineering values to verify the rationality of the model engineering parameters. The algorithm model is of good engineering application value. The research results can provide references for engineering investigation, design and construction parameter values, and also provide a new analysis method for geotechnical parameter value analyses.
-
表 1 各层泥炭质土的物理指标统计表
Table 1. Statistical table of physical indexes of peaty soil in each layer
地层编号 ρ
/(g·cm−3)Gs w
/%wL
/%wP
/%e a1-2
/MPawu
/%③11 1.61 2.30 110.1 135.4 100.0 2.49 1.62 24.33 ③2 1.42 2.42 101.2 119.2 87.2 2.40 1.32 18.36 ④1 1.25 1.95 145.7 182.1 146.4 2.93 1.49 40.10 ⑤12 1.27 1.97 130.6 174.3 140.6 2.52 1.12 38.77 表 2 不同物理指标间的相关性系数
Table 2. Correlation between different physical indicators
R,2 ρ GS w wL e a1-2 wP wu ρ 1.00 GS 0.64 1.00 w 0.87 0.55 1.00 wL 0.89 0.59 0.94 1.00 e 0.85 0.85 0.94 0.87 1.00 a1-2 0.62 0.42 0.81 0.68 0.85 1.00 wP 0.85 0.58 0.82 0.93 0.58 0.49 1.00 wu 0.83 0.66 0.73 0.78 0.70 0.45 0.80 1.00 表 3 各物理指标敏感度统计表
Table 3. Sensitivity statistics of each physical index
物理指标编码 X1(ρ) X2 (GS) X3 (w) X4(wL) X5(e) X6(a1-2) X7 (wP) X8 (wu) 敏感度 0.2349 0.2069 0.2540 0.2428 0.2608 0.2151 0.2410 0.2281 表 4 各组泥炭质土综合评价值排序
Table 4. Ranking of comprehensive evaluation of the peaty soil
组类 ρ/(g·cm−3) w/% GS e wL/% wP/% a1-2/MPa wu/% 综合评价 1 1.565 57.590 2.517 1.536 68.690 42.690 0.722 8.100 -1.295 2 1.533 67.980 2.397 1.565 71.730 44.910 0.812 12.670 -1.162 3 1.499 68.280 2.428 1.699 76.180 48.270 1.103 11.690 -1.096 4 1.483 72.120 2.436 1.824 84.800 49.560 0.977 12.270 -1.035 5 1.452 73.690 2.238 1.882 92.400 58.890 0.850 13.678 -0.928 6 1.454 78.070 2.426 1.971 90.320 53.550 1.121 12.100 -0.946 7 1.455 82.180 2.426 1.957 95.363 62.181 0.833 13.318 -0.900 … … … … … … … … … … 34 1.115 211.200 1.796 4.406 228.270 149.400 3.802 43.750 1.293 35 1.134 220.200 1.164 4.761 228.710 144.100 4.718 44.346 1.496 36 1.117 232.700 1.311 4.818 238.770 149.310 5.031 40.817 1.566 37 1.121 247.000 0.996 5.327 249.400 156.300 6.006 41.710 1.861 38 1.079 287.182 0.990 5.839 304.936 207.800 5.978 56.455 2.574 表 5 不同截尾法的原物理指标区间范围
Table 5. Original physical index range of different censoring methods
物理指标 均值μ 标准差σ 偏度系数c 变异系数δ 3σ区间范围 c33区间范围 c3区间范围 e 3.03 1.20 0.94 0.39 [0,6.63] [0.56,7.76] [0,7.76] w/% 133.02 62.09 0.93 0.46 [0,319.29] [4.48,377.03] [0,377.03] wL/% 148.63 61.26 0.87 0.41 [0,332.41] [18.14,385.71] [0,385.71] 表 6 物理指标样本检验结果
Table 6. Normal test result of the physical index sample
表 7 不同截尾法的新样本区间范围
Table 7. Range of new sample intervals for different censoring methods
物理指标 均值μ 标准差σ 偏度系数c 变异系数δ 3σ区间范围 c33区间范围 c3区间范围 e 2.19 0.40 0.39 0.18 [0.99,3.39] [1.45,3.55] [0.99,3.55] w/% 88.68 17.40 0.39 0.18 [36.48,140.88] [43.26,147.66] [36.48,140.88] wL/% 105.55 22.11 0.28 0.20 [39.22,171.88] [45.41,178.07] [39.22,178.08] ρ/(g·cm−3) 1.40 0.08 0.08 0.06 [1.13,1.66] [1.15,1.67] [1.13,1.67] Gs 2.25 0.45 -3.08 0.20 [0.89,3.61] [1.15,3.55] [1.15,3.61] a1-2/MPa 1.24 0.56 1.56 0.45 [0,2.93] [0.43,3.81] [0,2.93] wP/% 68.76 18.06 0.38 0.26 [14.57,122.97] [21.48,129.88] [14.57,129.88] wu/% 17.63 7.64 2.02 0.43 [0,40.57] [10.17,56.06] [0,56.06] 表 8 各层泥炭质土设计参数检验结果
Table 8. Test results of design parameters of the peaty soil in each layer
地层
编号ρ
/(g·cm−3)w
/%e IL a1-2
/MPa综合
评价所属类 ③11 1.37 103 2.41 0.15 1.63 −0.23 1 ③2 1.39 91.1 2.33 0.12 1.30 −0.26 1 ④1 1.20 142.3 2.92 −0.14 1.47 −0.08 1 ⑤12 1.26 127.7 2.53 −0.28 1.12 −0.16 1 -
[1] KOLAY P K, AMINUR M R, TAIB S N L, et al. Correlation between different physical and engineering properties of tropical peat soils from Sarawak[C]//GEO Shanghai International Conference 2010. June 3-5, 2010, Shanghai, China. Reston, VA, USA: American Society of Civil Engineers, 2010: 56-61. [2] PRICE J S,CAGAMPAN J,KELLNER E. Assessment of peat compressibility:is there an easy way?[J]. Hydrological Processes,2005,19(17):3469 − 3475. doi: 10.1002/hyp.6068 [3] SANTAGATA M,BOBET A,JOHNSTON C T,et al. One-dimensional compression behavior of a soil with high organic matter content[J]. Journal of Geotechnical and Geoenvironmental Engineering,2008,134(1):1 − 13. doi: 10.1061/(ASCE)1090-0241(2008)134:1(1) [4] 裴利华,杨醒宇,桂跃,等. 有机质含量及组分对泥炭土物理力学性质影响[J]. 水文地质工程地质,2022,49(2):77 − 85. [PEI Lihua,YANG Xingyu,GUI Yue,et al. Effect of Organic Matter Content and Composition on Physical and Mechanical Properties of Peat Soil[J]. Hydrogeology & Engineering Geology,2022,49(2):77 − 85. (in Chinese with English abstract) [5] 严春风,刘东燕,朱可善. 岩石抗剪强度指标互相关性的敏感度分析[J]. 水文地质工程地质,1997(6):45 − 47. [YAN Chunfeng,LIU Dongyan,ZHU Kesan. Sensitivity analysis of cross-correlation of rock shear strength index[J]. Hydrogeology & Engineering Geology,1997(6):45 − 47. (in Chinese with English abstract) [6] 程豪,田林. 昆明市某场地不同埋深下泥炭质土物理力学指标统计规律研究[J]. 贵州大学学报(自然科学版),2022,39(4):92 − 98. [CHENG Hao,TIAN Lin. Study on Statistical Law of Physical & Mechanical Indexes of Peat Soil under Different Buried Depths in a Site of Kunming City[J]. Journal of Guizhou University (Natural Science Eidition),2022,39(4):92 − 98. (in Chinese) [7] 丁祖德,付江,李夕松,等. 昆明泥炭质土的物理力学指标特征及相关性分析[J]. 公路工程,2018,43(4):86 − 91. [DING Zude,FU Jiang,LI Xisong,et al. Physical & mechanical index characteristics and correlation analysis of peaty soil in Kunming[J]. highway engineering,2018,43(4):86 − 91. (in Chinese) [8] 陈玉萍,袁志强,周博,等. 遗传算法优化BP网络在滑坡灾害预测中的应用研究[J]. 水文地质工程地质,2012,39(1):114 − 119. [CHEN Yuping,YUAN Zhiqiang,ZHOU Bo,et al. ] Application of BP Network Optimized by Genetic Algorithm in Landslide Disaster Prediction[J]. Hydrogeology & Engineering Geology,2012,39(1):114 − 119. (in Chinese with English abstract) [9] 马壮壮,束龙仓,季叶飞,等. 基于遗传算法的BP神经网络计算岩溶水安全开采量[J]. 水文地质工程地质,2016,43(1):22 − 27. [MA Zhuangzhuang,SHU Longcang,JI Yefei,et al. Calculation of Safe Extraction of Karst Water Based on BP Neural Network of Genetic Algorithm[J]. Hydrogeology & Engineering Geology,2016,43(1):22 − 27. (in Chinese with English abstract) [10] 阮永芬, 张虔, 乔文件, 等. 盾构施工地表沉降的无监督学习预估方法[J/OL]. 安全与环境学报: 1 − 11. [2022-09-05] DOI: 10.13637/j.issn.1009-6094.2022.0351. RUAN Yongfen, ZHANG Qian, QIAO Wenjian, et al. Unsupervised Learning Prediction Method of Ground Settlement in Shield Construction[J/OL]. Journal of Safety & Environment: 1 − 11. [2022-09-05]. DOI: 10.13637/j.issn.1009-6094.2022.0351. (in Chinese with English abstract) [11] 宫凤强,鲁金涛. 基于主成分分析与距离判别分析法的突水水源识别方法[J]. 采矿与安全工程学报,2014,31(2):236 − 242. [GONG Fengqiang,LU Jintao. Recognition method of mine water inrush sources based on the principal element analysis & distance discrimination analysis[J]. Journal of Mining & Safety Engineering,2014,31(2):236 − 242. (in Chinese with English abstract) [12] 任玉晓. 地震波速无监督深度学习反演方法及其在隧道超前探测中的应用[D]. 济南: 山东大学, 2021 REN Yuxiao. Unsupervised deep learning inversion method for seismic velocity & its application in tunnel forward prospecting[D]. Jinan: Shandong University, 2021. (in Chinese with English abstract) [13] 阮永芬,魏德永,高骏,等. K-means聚类分析高原湖相沉积软土参数[J]. 昆明理工大学学报(自然科学版),2020,45(1):85 − 91. [RUAN Yongfen,WEI Deyong,GAO Jun,et al. K-means Clustering Analysis of Soft Soil Parameters of Plateau Lake Sediment[J]. Journal of Kunming University of Science & Technology (Natural Science),2020,45(1):85 − 91. (in Chinese with English abstract) [14] 孔令奇,李翠娟. 岩土参数概率分布的系统推断研究[J]. 工业建筑,2022,52(1):129 − 136. [KONG Lingqi,LI Cuijuan. Study on System Inference of Probability Distribution of Geotechnical Parameters J]. industrial architecture,2022,52(1):129 − 136. (in Chinese with English abstract) [15] 殷瑞刚,魏帅,李晗,等. 深度学习中的无监督学习方法综述[J]. 计算机系统应用,2016,25(8):1 − 7. [Yin Ruigang,WEI Shuai,LI Han,et al. A Survey of Unsupervised Learning Methods in Deep Learning[J]. computer system application,2016,25(8):1 − 7. (in Chinese with English abstract) [16] 鲁金涛,李夕兵,宫凤强,等. 基于主成分分析与Fisher判别分析法的矿井突水水源识别方法[J]. 中国安全科学学报,2012,22(7):109 − 115. [LU Jintao,LI Xibing,Gong Fengqiang,et al. Identification method of mine water inrush source based on principal component analysis & Fisher discriminant analysis[J]. China Safety Science,2012,22(7):109 − 115. (in Chinese) doi: 10.3969/j.issn.1003-3033.2012.07.018 [17] 琚棋定,胡友彪,张淑莹. 基于主成分分析与贝叶斯判别法的矿井突水水源识别方法研究[J]. 煤炭工程,2018,50(12):90 − 94. [JU Qiding,HU Youbiao,ZHANG Shuying. Study on identification method of mine water inrush source based on principal component analysis & Bayes discriminant method[J]. Coal Engineering,2018,50(12):90 − 94. (in Chinese) [18] 孙吉贵,刘杰,赵连宇. 聚类算法研究[J]. 软件学报,2008,19(1):48 − 61. [SUN Jigui,LIU Jie,ZHAO Lianyu. Research on Clustering Algorithm[J]. Journal of Software,2008,19(1):48 − 61. (in Chinese with English abstract) doi: 10.3724/SP.J.1001.2008.00048 [19] 张麟, 潘红岩. 聚类分析算法应用研究[J]. 数字技术与应用, 2016(10) ZHANG Lin, PAN Hongyan. Application Research of Clustering Analysis Algorithm [J]. Digital Technology & Application, 2016(10). (in Chinese) [20] 刘荣. 人工神经网络基本原理概述[J]. 计算机产品与流通, 2020(6) LIU Rong. Basic Principle of Artificial Neural Network[J]. Computer Products & Circulation, 2020(6). (in Chinese ) [21] 袁冰清,程功,郑柳刚. BP神经网络基本原理[J]. 数字通信世界,2018(8):28 − 29. [YUAN Bingqing,CHENG Gong,ZHENG Liugang. Basic principle of BP neural network[J]. Digital Communication World,2018(8):28 − 29. (in Chinese with English abstract) [22] 刘伟, 赵福玉, 杨文辉, 等. 安嵩线草海段泥炭质土的特征及性质[J]. 岩土工程学报, 2013, 35(增刊2): 671 − 674 LIU Wei, ZHAO Fuyu, YANG Wenhui, et al. Characteristics and Properties of Peaty Soil in Caohai Section of Ansong Line[J]. Chinese Journal of Geotechnical Engineering. 2013, (Sup 2): 671 − 674. (in Chinese ) [23] 阮永芬,刘岳东,王东,等. 昆明泥炭与泥炭质土对建筑地基的影响[J]. 昆明理工大学学报(理工版),2003,28(3):121 − 124. [RUAN Yongfen,LIU Yuedong,WANG Dong,et al. Influence of peat & peaty soil on building foundation in Kunming[J]. Journal of Kunming University of Science & Technology,2003,28(3):121 − 124. [24] 柴春阳,张广泽,胡婷. 丽香铁路冰水沉积型泥炭土特性及工程对策分析[J]. 地下空间与工程学报,2017,13(增刊 2):757 − 761. [CHAI Chunyang,ZHANG Guangze,HU Ting. Analysis on characteristics and engineering countermeasures of ice water sedimentary peat soil in Lixiang Railway[J]. Underground Space & Engineering Journal,2017,13(Sup 2):757 − 761. (in Chinese) [25] 李强,李刚. 沪苏浙高速公路泥炭质土的工程性质研究[J]. 公路工程,2010,35(06):99 − 103. [LI Qiang,LI Gang. Study on Engineering Properties of Peat Soil in Hu-Su-Zhe Expressway[J]. Highway Engineering,2010,35(06):99 − 103. doi: 10.3969/j.issn.1674-0610.2010.06.026 [26] 沈俊海,王艳群. 桓仁至永陵高速公路泥炭质土工程地质评价[J]. 北方交通,2009(02):57 − 59. [SHEN Junhai,WANG Yanqun. Evaluation of peaty soil engineering geology of Huanren-Yongling Expressway[J]. North Traffic,2009(02):57 − 59. (in Chinese) doi: 10.3969/j.issn.1673-6052.2009.02.020 [27] 刘月华,阮永芬,施建伟,等. 深厚泥炭质土场地深基坑工程事故分析[J]. 工业安全与环保,2022,48(3):33 − 37. [LIU Yuehua,RUAN Yongfen,SHI Jianwei,et al. Accident analysis of deep foundation pit engineering in deep peat soil site[J]. Industrial safety & environmental protection,2022,48(3):33 − 37. doi: 10.3969/j.issn.1001-425X.2022.03.008 [28] 谢俊文. 泥炭的工程地质特性[J]. 工程勘察,1981(5):31 − 34. [XIE Junwen. Engineering geological characteristics of peat[J]. Engineering investigation,1981(5):31 − 34. [29] 宫凤强,黄天朗,李夕兵. 岩土抗剪强度参数的最优概率分布函数推断方法[J]. 岩土工程学报,2016,38(增刊 2):204 − 209. [GONG Fengqiang,HUANG Tianlang,LI Xibing. Inference method for optimal probability distribution function of shear strength parameters in geotechnical engineering[J]. Chinese Journal of Geotechnical Engineering,2016,38(Sup 2):204 − 209. (in Chinese with English abstract) [30] 宫凤强,侯尚骞,李夕兵. 岩土参数截尾分布的正态信息扩散推断方法[J]. 武汉大学学报(工学版),2016,49(5):661 − 667. [GONG Fengqiang,HOU Shangqian,LI Xibing. Normal information diffusion inference method for truncated distribution of geotechnical parameters[J]. Engineering Journal of Wuhan University,2016,49(5):661 − 667. (in Chinese with English abstract) -