基于MetaOD模型选择的岩土工程数据异常检测方法

邹彤彤; 刘孝义; 刘金泉; 袁海梁; 卢玉斌; 张万虎

doi:10.19509/j.cnki.dzkq.2022.0041

基于MetaOD模型选择的岩土工程数据异常检测方法

doi: 10.19509/j.cnki.dzkq.2022.0041

邹彤彤^{1, 2,},
刘孝义³,
刘金泉^2, ,,
袁海梁³,
卢玉斌²,
张万虎³

基金项目:

国家自然科学基金青年基金项目 51809253

福建省自然科学基金项目 2019J01142

详细信息

作者简介:
邹彤彤(1996—)，女，现正攻读应用统计专业硕士学位，主要从事岩土工程中数据异常检测工作。E-mail: 17320026010@163.com

通讯作者:
刘金泉(1989—)，男，副研究员，主要从事岩土工程教学与科研工作。E-mail: jinquanliu99@163.com

中图分类号: X830.3
计量
- 文章访问数: 854
- PDF下载量: 47
- 被引次数: 0
出版历程
- 收稿日期: 2021-05-27

Outlier detection method for geotechnical engineering based on MetaOD model selection

摘要

摘要: 岩土工程现场及室内参数测试数据是工程施工、设计、评价的基础。异常数据的存在往往会误导施工、设计等参数的确定, 数据异常检测是确保工程安全可靠的最基本但极为重要的工作。针对传统异常检测算法没有模型选择这一过程而导致检测的盲目性, 提出了基于元学习的异常检测算法(meta-learning outlier detection, MetaOD)和数据挖掘算法相结合的异常检测模型体系。该体系首先根据数据的特点选择适合不同数据类型的初始模型类型及其参数, 并对选择出的同类型算法的参数进行求均值处理; 然后再采用遴选出的算法进行数据异常诊断, 进而提高异常检测的准确性。为了评估模型的有效性, 采用加州大学欧文分校提出的机器学习检验数据集(glass数据集)进行检验分析。结果显示, 采用该模型体系进行异常检测时查准率达到96.41%, 远高于其他检测算法。最后, 应用该模型体系对澳门花岗岩单轴抗压强度数据集和均昌隧道的地下水位监测数据进行了异常检测分析, 并分别识别出9个和10个异常点。
- 岩土工程异常检测 /
- MetaOD算法 /
- 模型选择 /
- 数据挖掘
Abstract: The geotechnical engineering field and indoor parameter test data are the foundation of engineering construction, design and evaluation. The existence of abnormal data often misleads the determination of parameters such as construction and design. Data anomaly detection is the most basic but extremely important task to ensure the safety and reliability of a project. Aiming at the blindness of detection due to the lack of model selection in traditional anomaly detection algorithms, this paper proposes an anomaly detection model system based on a combination of meta-learning outlier detection (MetaOD) and data mining algorithms. The system first selects the initial model class and its parameters suitable for different data types according to the characteristics of the data, averages the selected parameters of the same type of algorithm, and then uses the selected algorithm to diagnose data anomalies, thereby improving the anomaly accuracy of detection. To evaluate the effectiveness of the model, the machine learning test dataset (glass dataset) proposed by the University of California Irvine, is used for test analysis. The results show that the accuracy rate of anomaly detection using this model system reaches 96.41%, which is much higher than that of other detection algorithms. Finally, the model system is applied to the uniaxial compressive strength dataset of the Macau granite and the groundwater monitoring data of the Junchang Tunnel to carry out anomaly detection and analysis and to identify 9 and 10 abnormal points, respectively.
- geotechnical engineering anomaly detection /
- MetaOD algorithm /
- model selection /
- data mining

HTML全文

图 1 流程图

Figure 1. Flow chart

下载: 全尺寸图片幻灯片

图 2 各变量分布及相关性可视化图

Figure 2. Visualization of the distribution and correlation of each variable

下载: 全尺寸图片幻灯片

图 3 本次变化量和累计变化量时序图

图中63, 1 440分别代表监测组编号及地下水位(m)

Figure 3. Time sequence diagram of current change and cumulative change

下载: 全尺寸图片幻灯片

表 1 分类结果混淆矩阵

Table 1. Confusion matrix of classification results

真实情况	预测结果
真实情况	正例	反例
正例	TP(真正例)	FN(假反例)
反例	FP(假正例)	TN(真反例)

下载: 导出CSV

表 2 glass数据模型选择结果

Table 2. The selection results of the glass data model

排名	模型	参数(领域数)
1	ABOD	5
2	ABOD	15
3	ABOD	20
4	ABOD	25
5	IForest	(20，0.2)

下载: 导出CSV

表 3 glass数据集在不同参数下的混淆矩阵和查准率

Table 3. Confusion matrix and precision of the glass dataset under different parameters

模型	TP	FN	FP	TN	查准率/%	运算时间/s
ABOD(5)	187	18	7	2	96.39	0.100 8
ABOD(15)	186	19	7	2	96.37	0.435 9
ABOD(20)	187	18	7	2	93.39	0.770 0
ABOD(25)	186	19	6	3	96.37	1.364 4
ABOD(16)	188	17	7	2	96.41	0.484 9

下载: 导出CSV

表 4 该模型算法与常规算法的检测结果比较

Table 4. Comparison of the detection results of the model algorithm and the conventional algorithm

模型	TP	FN	FP	TN	查准率/%	运算时间/s
ABOD (16)	188	17	7	2	96.41	0.484 9
COF	185	20	7	2	90.24	0.195 8
HBOS	184	21	8	1	89.76	1.303 5
OCSVM	184	21	8	1	89.76	0.014 8
LODA	184	21	8	1	89.76	0.027 4
CBLOF	185	20	7	2	90.24	1.096 7
COPOD	184	21	8	1	89.76	0.036 9
MCD	183	22	9	0	89.27	0.060 5
PCA	184	21	8	1	89.76	0.012 6
IForest	185	20	7	2	90.24	0.171 3

下载: 导出CSV

表 5 Ⅲ级花岗岩数据各变量间相关性

Table 5. Correlation among variables of the Macau Ⅲ-level granite dataset

	UCS	I_S 50	R_L	v_p	η_e	G_s
UCS	1	0.75	0.66	0.77	-0.56	0.48
I_S 50		1	0.65	0.46	-0.50	0.41
R_L			1	0.50	-0.68	0.57
v_p				1	-0.44	0.37
η_e					1	0.91
G_s						1
注：UCS.单轴抗压强度；I_S 50.点荷载指数；R_L.施密特锤回弹值；v_p.纵波波速；η_e.有效孔隙率；G_s.相对密度

下载: 导出CSV

表 6 Ⅲ级花岗岩数据模型选择结果

Table 6. Selection results of the Macau Ⅲ-level granite dataset model

排名	模型	参数
1	HBOS	(5，0.1)
2	IForest	(20，0.5)
3	IForest	(75，0.3)
4	IForest	(200，0.2)
5	IForest	(20，0.2)

下载: 导出CSV

表 7 均昌隧道的地下水位监测数据模型选择结果

Table 7. Selected results of the groundwater monitoring data model for the Junchang Tunnel

排名	模型	参数
1	LODA	(25, 200)
2	LODA	(15, 20)
3	IForest	(50, 0.2)
4	LODA	(25, 50)
5	LODA	(30, 20)

下载: 导出CSV

参考文献(28)

[1]	Bieniawski Z T. The point-load test in geotechnical practice[J]. Engineering Geology, 1975, 9(1): 1-11. doi: 10.1016/0013-7952(75)90024-1
[2]	Cargill J S, Shakoor A. Evaluation of empirical methods for measuring the uniaxial compressive strength of rock[J]. International Journal of Rock Mechanics and Mining Sciences & Geomechanics Abstracts. Pergamon, 1990, 27(6): 495-503. https://www.sciencedirect.com/science/article/pii/014890629091001N
[3]	柴波, 陶阳阳, 杜娟, 等. 基于Hoek-Brown准则的节理岩体能量参数估算[J]. 地质科技通报, 2020, 39(1): 78-85. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202001010.htm Chai B, Tao Y Y, Du J, et al. Energetics parameter estimation of jointed rock mass based on Hoek-Brown failure criterion[J]. Bulletin of Geological Scienceand Technology, 2020, 39(1): 78-85(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202001010.htm
[4]	李宗强, 李居铜, 张爱勤, 等. 土木工程试验方法与数据处理[M]. 哈尔滨: 哈尔滨工业大学出版社, 2014. Li Z Q, Li J T, Zhang A Q, et al. Civil engineering test methods and data processing[M]. Harbin: Harbin Institute of Technology Press, 2014(in Chinese).
[5]	马祥配, 刘福臣. 岩土工程检测常见问题处理[J]. 中国勘察设计, 2013(3): 96-98. https://www.cnki.com.cn/Article/CJFDTOTAL-KCSJ201303031.htm Ma X P, Liu F C. Handling of common problems in geotechnical engineering inspection[J]. China Engineering Consulting, 2013(3): 96-98(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-KCSJ201303031.htm
[6]	杨威. 郴宁高速公路万华岩边坡监测与稳定性评价方法研究[D]. 长沙: 中南大学, 2013. Yang W. Study of monitoring methods and stability evaluation method of Wanhuayan slope in Chenning highway[D]. Changsha: Central South University, 2013(in Chinese).
[7]	李明. 盾构隧道长期健康监测与评价[D]. 北京: 中国科学院大学, 2015. Li M. Long-term health monitoring and evaluation of shield tunnels[D]. Beijing: University of Chinese Academy of Sciences, 2015(in Chinese).
[8]	刘鸿飞, 刘俊芳, 苏跃宏, 等. 无侧限抗压强度异常值处理新方法的研究[J]. 岩土工程学报, 2020, 42(增刊1): 137-140. https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC2020S1027.htm Liu H F, Liu J F, Su Y H, et al. New method for dealing with unconfined compressive strength outliers[J]. Chinese Journal of Geotechnical Engineering, 2020, 42(S1): 137-140(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-YTGC2020S1027.htm
[9]	Mu H Q, Yuen K V. Novel outlier-resistant extended Kalman filter for robust online structural identification[J]. Journal of Engineering Mechanics, 2015, 141(1): 04014100. doi: 10.1061/(ASCE)EM.1943-7889.0000810
[10]	Ramaswamy S, Rastogi R, Shim K. Efficient algorithms for mining outliers from large data sets[C]//Anon. Proceedings of the 2000 ACM SIGMOD International Conference on Management of Data. [S.l.]: [s.n.], 2000: 427-438.
[11]	Breunig M M, Kriegel H P, Ng R T, et al. LOF: Identifying density-based local outliers[C]//Anon. Proceedings of the 2000 ACM SIGMOD International Conference on Management of Data. [S.l.]: [s.n.], 2000: 93-104.
[12]	谭文侃, 叶义成, 胡南燕, 等. LOF与改进SMOTE算法组合的强烈岩爆预测[J]. 岩石力学与工程学报, 2021, 40(6): 1186-1194. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202106010.htm Tan W K, Ye Y C, Hu N Y, et al. Severe rock burst prediction based on the combination of LOF and improved SMOTE algorithm[J]. Chinese Journal of Rock Mechanics and Engineering, 2021, 40(6): 1186-1194(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202106010.htm
[13]	He Z, Xu X, Deng S. Discovering cluster-based local outliers[J]. Pattern Recognition Letters, 2003, 24(9/10): 1641-1650.
[14]	Tang J, Chen Z, Fu A W C, et al. Enhancing effectiveness of outlier detections for low density patterns[C]//Anon. Pacific-Asia Conference on Knowledge Discovery and Data Mining. Berlin, Heidelberg: Springer, 2002: 535-548.
[15]	Papadimitriou S, Kitagawa H, Gibbons P B, et al. Loci: Fast outlier detection using the local correlation integral[C]//Anon. Proceedings 19th International Conference on Data Engineering (Cat. No. 03CH37405). [S.l.]: IEEE, 2003: 315-326.
[16]	Liu F T, Ting K M, Zhou Z H. Isolation forest[C]//Anon. 2008 Eighthieee International Conference on Data Mining. [S.l.]: IEEE, 2008: 413-422.
[17]	Zhao Y, Rossi R A, Akoglu L. Automating outlier detection via meta-learning[J]. arXiv preprint arXiv: 2009.10606, 2020.
[18]	Wolpert D H, Macready W G. No free lunch theorems for optimization[J]. IEEE Transactions on Evolutionary Computation, 1997, 1(1): 67-82. doi: 10.1109/4235.585893
[19]	Bache K, Lichman M. UCI Machine learning repository (2013)[EB/OL]. [2013-03-27](2021-05-10). Available: http://archive.ics.uci.edu/ml
[20]	Chen L, Guo Z, Yin K, et al. The influence of land use and land cover change on landslide susceptibility: a case study in Zhushan Town, Xuan′en County (Hubei, China)[J]. Natural hazards and earth system sciences, 2019, 19(10): 2207-2228. doi: 10.5194/nhess-19-2207-2019
[21]	周志华. 机器学习[M]. 北京: 清华大学出版社, 2016. Zhou Z H. Machine learning[M]. Beijing: Tsinghua University Press, 2016(in Chinese).
[22]	Kriegel H P, Schubert M, Zimek A. Angle-based outlier detection in high-dimensional data[C]//Anon. Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. [S.l.]: [s.n.], 2008: 444-452.
[23]	柏道远, 李彬, 李银敏, 等. 湖南常德-安仁断裂印支期构造运动分段性: 来自花岗岩的约束[J]. 地质科技通报, 2021, 40(5): 173-187. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202105019.htm Bai D Y, Li B, Li Y M, et al. Segmentation of the movement in Indosinian of the Changde-Anren fault in Hunan: Constraints from granite[J]. Bulletin of Geological Science and Technology, 2021, 40(5): 173-187(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202105019.htm
[24]	Ng I T, Yuen K V, Lau C H. Predictive model for uniaxial compressive strength for Grade Ⅲ granitic rocks from Macao[J]. Engineering Geology, 2015, 199: 28-37. doi: 10.1016/j.enggeo.2015.10.008
[25]	姚文礼. 四川盆地须家河组致密砂岩物源体系的控储作用[J]. 地质科技通报, 2021, 40(5): 223-230. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202105023.htm Yao W L. Reservoir control of tight sandstone provenance system in Xujiahe formation, Sichuan Basin[J]. Bulletin of Geological Science and Technology, 2021, 40(5): 223-230(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202105023.htm
[26]	Goldstein M, Dengel A. Histogram-based outlier score (hbos): A fast unsupervised anomaly detection algorithm[C]//Anon. KI-2012: Poster and demo track. [S.l.]: [s.n.], 2012: 59-63.
[27]	江欣悦, 李静, 郭林, 等. 豫北平原浅层地下水化学特征与成因机制[J]. 地质科技通报, 2021, 40(5): 290-300. https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202105030.htm Jiang X Y, Li J, Guo L, et al. Chemical characteristics and formation mechanism of shallow ground water in the northern Henan Plain[J]. Bulletin of Geological Scienceand Technology, 2021, 40(5): 290-300(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DZKQ202105030.htm
[28]	Pevn T. Loda: Lightweight on-line detector of anomalies[J]. Machine Learning, 2016, 102(2): 275-304. doi: 10.1007/s10994-015-5521-0