基于SSA-Adam-BP神经网络模型的堰塞坝稳定性预测

宋宜祥; 张晓波; 黄达

doi:10.19509/j.cnki.dzkq.2022.0040

基于SSA-Adam-BP神经网络模型的堰塞坝稳定性预测

doi: 10.19509/j.cnki.dzkq.2022.0040

基金项目:

国家自然科学基金项目 41902290

国家自然科学基金项目 41672300

国家自然科学基金项目 41972297

河北省自然科学基金项目 D2020202002

河北省自然科学基金项目 D202102002

详细信息

作者简介:
宋宜祥(1987—)，男，讲师，硕士生导师，主要从事地质灾害和岩土工程数值模拟研究。E-mail: syxdlut2010@163.com

通讯作者:
黄达(1976—)，男，教授，博士生导师，主要从事岩石力学与地质灾害方面研究。E-mail: dahuang@hebut.edu.cn

中图分类号: P642.2
计量
- 文章访问数: 516
- PDF下载量: 45
- 被引次数: 0
出版历程
- 收稿日期: 2021-05-10

Stability prediction of landslide dams based on SSA-Adam-BP neural network model

摘要

摘要: 现有的堰塞坝稳定性预测模型多为线性模型, 无法充分考虑堰塞坝稳定性与其形态特征和水域条件之间的复杂非线性关系。鉴于此, 结合反向传播神经网络模型和樽海鞘优化算法, 提出了一种新型的堰塞坝稳定性预测模型SSA-Adam-BP。该模型通过网格搜索法选取确定模型结构的最佳超参数组合, 进而利用交叉验证和绘制ROC曲线的方式分别对采用不同优化算法的模型进行评估。使用开源数据库中的全球153例堰塞坝数据对模型的实际应用进行了说明及验证。与传统线性模型的对比表明神经网络模型预测准确率较高, 具有较低的误报率。将SSA与Adam优化算法结合提高了BP模型的全局搜索能力, 其平均交叉验证准确率达到了91.73%, 能够使用较少的参数实现对堰塞坝稳定性快速准确的预测。SSA-Adam-BP模型对近年来典型工程的稳定性能够准确预测, 具有一定的实用性和系统平台推广应用价值。
- 堰塞坝稳定性 /
- 机器学习 /
- 稳定性预测 /
- BP神经网络 /
- SSA优化算法
Abstract: Most of the existing landslide dam stability prediction models are linear models, which cannot fully consider the complex nonlinear relationship between landslide dam stability and its morphological characteristics and hydrodynamic conditions.In view of this, a new SSA-Adam-BP model for predicting the stability of landslide dams is proposed by combining the back propagation neural network model and the salp optimization algorithm.The grid search method is used to select the best combination of hyperparameters that can determine the structure of the model.Then, the models with different optimization algorithms are evaluated by cross-validation and ROC curve drawing.The practical application of the model is explained and verified by using the global data of 153 landslide dams in the open source database.Compared with the traditional linear model, the combination of the SSA and Adam optimization algorithm improves the global search ability of the BP model, and its average cross-verification accuracy reaches 91.73%.It not only has a lower misjudgment rate but can also use fewer parameters to quickly and accurately predict the stability of landslide dams.The SSA-Adam-BP model can accurately predict the stability of typical projects in recent years, with certain practicality and system platform promotion application value.
- landslide dam stability /
- machine learning /
- stability prediction /
- BP neural network /
- SSA optimization algorithm

HTML全文

图 1 单隐藏层BP神经网络结构图

Figure 1. BP neural network structure of single hidden layer

下载: 全尺寸图片幻灯片

图 2 SSA-Adam-BP神经网络训练流程图

Figure 2. Flow chart of SSA-Adam-BP neural network training

下载: 全尺寸图片幻灯片

图 3 采用不同处理方式的变量分布示意图

Figure 3. Schematic diagrams of variable distribution using different processing methods

下载: 全尺寸图片幻灯片

图 4 不同BP神经网络模型的ROC曲线及AUC值

Figure 4. ROC curve and AUC value of different BP neural network models

下载: 全尺寸图片幻灯片

图 5 预测失误案例位置分布图

Figure 5. Location distribution map of prediction error cases

下载: 全尺寸图片幻灯片

图 6 中国32例堰塞坝的稳定性预测结果

Figure 6. Stability prediction results of 32 landslide dams in China

下载: 全尺寸图片幻灯片

表 1 Ermini堰塞坝数据库

Table 1. Landslide dam database established by Ermini

序号	坝体高度H/m	坝体体积V/m³	集水区面积A/km²	稳定性分类	参考文献
1	20	1 500 000	5	SD	[10]
2	15	400 000	15	SD	[10]
⋮	⋮	⋮	⋮	⋮	⋮
84	90	18 000 000	147	UD	[10]

下载: 导出CSV

表 2 Tabata堰塞坝数据库

Table 2. Landslide dam database established by Tabata

序号	坝体高度H/m	坝体体积V/m³	集水区面积A/km²	稳定性分类	参考文献
1	10	180 000	60	UD	[26]
2	20	100 000	83	UD	[26]
⋮	⋮	⋮	⋮	⋮	⋮
37	20	22 000	14.5	SD	[26]

下载: 导出CSV

表 3 121例堰塞坝数据基本统计信息

Table 3. Basic statistical information of 121 cases based on the landslide dam database

特征变量	坝体高度H/m	坝体体积V/m³	集水区面积A/km²
最小值	3.00	8.70×10⁷	0.19
最大值	500.00	2.20×10⁹	4.50×10⁴
平均值	52.88	5.70×10⁷	731.61
标准差	65.87	2.86×10⁸	4 321.71

下载: 导出CSV

表 4 Shapiro-Wilk检验结果

Table 4. Testing results of Shapiro-Wilk

变量	统计量W	p值	在5%水平下的结论
H′	0.68	0.00	排除正态性
V′	0.19	0.00	排除正态性
A′	0.14	0.00	排除正态性
lg(H′)	0.99	0.79	不能排除正态性
lg(V′)	0.999	0.70	不能排除正态性
lg(A′)	0.99	0.11	不能排除正态性

下载: 导出CSV

表 5 不同BP神经网络模型的交叉验证准确率

Table 5. Cross-verification accuracy of different BP neural network models

模型	准确率/%					平均准确率/%
模型	K=1	K=2	K=3	K=4	K=5	平均准确率/%
Adam-BP	87.50	87.50	91.67	91.67	88.00	89.27
SSA-BP	87.50	83.33	91.67	91.67	88.00	88.43
SSA-Adam-BP	91.67	87.50	95.83	91.67	92.00	91.73

下载: 导出CSV

表 6 堰塞坝稳定性线性预测模型

Table 6. Linear prediction model of landslide dam stability

提出者	判别公式	判别标准
Ermini等^[10]	$DBI = {\rm{lg}}\left( {\frac{{A \times H}}{V}} \right)$	DBI < 2.75稳定； 2.75 < DBI < 3.08 无法确定；DBI>3.08 不稳定
Dong等^[27]	$\begin{array}{c} D = - 2.13{\rm{lg}}\left( A \right) - \\ 4.08{\rm{lg}}\left( H \right) + \\ 2.94{\rm{lg}}\left( V \right) + 4.09 \end{array}$	D>0稳定; D≤0不稳定
Dong等^[28]	$\begin{array}{c} {L_{\rm{s}}} = - 4.48{\rm{lg}}\left( A \right) - \\ 9.31{\rm{lg}}\left( H \right) + \\ 6.61{\rm{lg}}\left( V \right) + 6.39\\ {P_{\rm{s}}} = \frac{1}{{1 + {{\rm{e}}^{ - {L_{\rm{s}}}}}}} \end{array}$	P_s>0.5稳定; P_s≤0.5不稳定

下载: 导出CSV

表 7 各堰塞坝稳定性预测模型的预测结果

Table 7. Prediction results of the stability prediction models of landslide dams

模型	准确数量	准确率/%	误报数量	误报率/%
DBI	23	62.16	7	18.92
AHV_Dis	26	70.27	6	16.22
AHV_Log	27	72.97	5	13.51
SSA-Adam-BP	32	86.49	1	2.70

下载: 导出CSV

表 8 各输入参数的平均影响值

Table 8. Mean influence value of each input parameter

影响因素	Y₁	Y₂	平均影响值
坝体高度	32.04	46.31	-0.17
坝体体积	55.54	23.56	0.38
集水区面积	31.02	51.27	-0.24

下载: 导出CSV

表 9 中国近年来典型堰塞坝事件^[31]

Table 9. Typical landslide dam cases in China in recent years

序号	形成时间	名称	坝体高度H/m	坝体体积V/m³	集水区面积A/km²	稳定性分类
1	2000	易贡	80	300 000 000	13 533	不稳定
2	2008	白果村	15	400 000	3 564	不稳定
⋮	⋮	⋮	⋮	⋮	⋮	⋮
32	2018	白格(2)	81	10 000 000	173 484	不稳定

下载: 导出CSV

参考文献(35)

[1]	年廷凯, 吴昊, 陈光齐, 等. 堰塞坝稳定性评价方法及灾害链效应研究进展[J]. 岩石力学与工程学报, 2018, 37(8): 1796-1812. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201808003.htm Nian T K, Wu H, Chen G Q, et al. Research progress on stability evaluation method and disaster chain effect of landslide dam[J]. Chinese Journal of Rock Mechanics and Engineering, 2018, 37(8): 1796-1812(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201808003.htm
[2]	Zhong Q M, Chen S S, Mei S A, et al. Numerical simulation of landslide dam breaching due to overtopping[J]. Landslides, 2018, 15(6): 1183-1192. doi: 10.1007/s10346-017-0935-3
[3]	Costa J E, Schuster R L. The formation and failure of natural dams[J]. Geological Society of America Bulletin, 1988, 100(7): 1054-1068. doi: 10.1130/0016-7606(1988)100<1054:TFAFON>2.3.CO;2
[4]	石振明, 马小龙, 彭铭, 等. 基于大型数据库的堰塞坝特征统计分析与溃决参数快速评估模型[J]. 岩石力学与工程学报, 2014, 33(9): 1780-1790. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201409008.htm Shi Z M, Ma X L, Peng M, et al. Statistical analysis of dam characteristics based on large-scale database and rapid evaluation model of collapse parameters[J]. Chinese Journal of Rock Mechanics and Engineering, 2014, 33(9): 1780-1790(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX201409008.htm
[5]	叶华林. 基于堰塞坝几何形态的数理统计对其稳定性影响研究[D]. 成都: 西南交通大学, 2018. Ye H L. Research on the influence of mathematical statistics based on the geological characteristics of dam dams to its stability[D]. Chengdou: Southwest Jiaotong University, 2018(in Chinese with English abstract).
[6]	FanX M, Dufresne A, Subramanian S S, et al. The formation and impact of landslide dams: State of the art[J]. Earth-Science Reviews, 2020, 203: 103116. doi: 10.1016/j.earscirev.2020.103116
[7]	彭铭, 王开放, 张公鼎, 等. 堰塞坝溃坝模型实验研究综述[J]. 工程地质学报, 2020, 28(5): 1007-1015. https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202005010.htm Peng M, Wang K F, Zhang G D, et al. Review of model experimental studies on break of landslide dams[J]. Journal of Engineering Geology, 2020, 28(5): 1007-1015(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-GCDZ202005010.htm
[8]	Costa J E, Schuster R L. Documented historical landslide dams from around the world[R]. : US Geological Survey Open-file Report, 1991: 91-239.
[9]	Casagli N, Ermini L. Geomorphic analysis of landslide dams in the Northern Apennine[J]. Chikei, 1999, 20(3): 219-249.
[10]	Ermini L, Casagli N. Prediction of the behaviour of landslide dams using a geomorphological dimensionless index[J]. Earth Surface Processes & Landforms, 2003, 28(1): 31-47.
[11]	Korup O. Geomorphometric characteristics of New Zealand landslide dams[J]. Engineering Geology, 2004, 73(1/2): 13-35.
[12]	Stefanelli C T, Catani F, Casagli N. Geomorphological investigations on landslide dams[J]. Geoenvironmental Disasters, 2015, 2(1): 1-15. doi: 10.1186/s40677-014-0008-z
[13]	Stefanelli C T, Segoni S, Casagli N, et al. Geomorphic indexing of landslide dams evolution[J]. Engineering Geology, 2016, 208: 1-10. doi: 10.1016/j.enggeo.2016.04.024
[14]	Shan Y B, Chen S S, Zhong Q M. Rapid prediction of landslide dam stability using the logistic regression method[J]. Landslides, 2020, 17(12): 2931-2956. doi: 10.1007/s10346-020-01414-6
[15]	罗丹, 李昌彩, 吴长彬. 基于微粒群-BP神经网络算法的堆石坝坝体变形监控模型研究[J]. 岩石力学与工程学报, 2012, 31(增刊1): 2926-2931. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX2012S1043.htm Luo D, Li C C, Wu C B. Study of monitoring model of faced rockfill dam's deformation based on particle swarm optimization-bp algorithm[J]. Chinese Journal of Rock Mechanics and Engineering, 2012, 31(S1): 2926-2931(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX2012S1043.htm
[16]	郭子正, 殷坤龙, 付圣, 等. 基于GIS与WOE-BP模型的滑坡易发性评价[J]. 地球科学, 2019, 44(12): 4299-4312. https://www.cnki.com.cn/Article/CJFDTOTAL-DQKX201912040.htm Guo Z Z, Yin K L, Fu S, et al. Evaluation of landslide susceptibility based on GIS and WOE-BP model[J]. Earth Science, 2019, 44(12): 4299-4312(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DQKX201912040.htm
[17]	马瑶, 赵江南. 机器学习方法在矿产资源定量预测应用研究进展[J]. 地质科技通报, 2021, 40(1): 132-141. doi: 10.19509/j.cnki.dzkq.2021.0108 Ma Y, Zhao J N. Advances in the application of machine learning methods in mineral prospectivity mapping[J]. Bulletin of Geological Science and Technology, 2021, 40(1): 132-141(in Chinese with English abstract). doi: 10.19509/j.cnki.dzkq.2021.0108
[18]	焦敬品, 李勇强, 吴斌, 等. 基于BP神经网络的管道泄漏声信号识别方法研究[J]. 仪器仪表学报, 2016, 37(11): 2588-2596. doi: 10.3969/j.issn.0254-3087.2016.11.023 Jiao J P, Li Y Q, Wu B, et al. Research on identi-fication method of pipeline leakage sound signal based on bpneural network[J]. Journal of Instrumentation, 2016, 37(11): 2588-2596(in Chinese with English abstract). doi: 10.3969/j.issn.0254-3087.2016.11.023
[19]	Hecht-Nielsen R. Theory of the backpropagation neural network[J]. Neural Networks, 1988, 1: 593-605.
[20]	Guo Z Z, Chen L X, Gui L, et al. Landslide displacement prediction based on variational mode decomposition and WA-GWO-BP model[J]. Landslides, 2020, 17: 567-583. doi: 10.1007/s10346-019-01314-4
[21]	Mirjalili S, Gandomi A H, Mirjalili S Z, et al. Salp swarm algorithm: A bio-inspired optimizer for engineering design problems[J]. Advances in Engineering Software, 2017, 114: 163-191. doi: 10.1016/j.advengsoft.2017.07.002
[22]	刘小龙, 许岩, 徐维军. 基于统计引导和多项式差分学习的樽海鞘优化算法[J]. 运筹与管理, 2021, 30(1): 43-49. https://www.cnki.com.cn/Article/CJFDTOTAL-YCGL202101008.htm Liu X L, Xu Y, Xu W J. Salp swarm algorithm based on statistical guidance and polynomial difference learning[J]. Operations Research and Management Science, 2021, 30(1): 43-49(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-YCGL202101008.htm
[23]	杨晓帆, 陈廷槐. 人工神经网络固有的优点和缺点[J]. 计算机科学, 1994, 21(2): 23-26. https://www.cnki.com.cn/Article/CJFDTOTAL-JSJA199402004.htm Yang X F, Chen T H. Inherent advantages and disadvantages of artificial neural networks[J]. Computer Science, 1994, 21(2): 23-26(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-JSJA199402004.htm
[24]	Kingma D, Ba J. Adam: A method for stochastic optimization[C]//International Conference Leaning Representation. :, 2015: 1-13.
[25]	万里明, 吴均, 卢军凯, 等. 基于Adam-神经网络的致密砂岩脆性评价方法: 以南堡凹陷高北边坡为例[J]. 地质科技通报, 2020, 39(2): 94-103. doi: 10.19509/j.cnki.dzkq.2020.0210 Wan L M, Wu J, Lu J K, et al. Brittleness evaluation method of tight sandstone based on Adam-neural network: A case study of a block in Gaobei Slope, Nanpu Sag[J]. Bulletin of Geological Science and Technology, 2020, 39(2): 94-103(in Chinese with English abstract). doi: 10.19509/j.cnki.dzkq.2020.0210
[26]	Tabata S, Mizuyama T, Inoue K. Natural landslide dams hazards[J]. Engineering Geology, 2002, 110(3): 162-171.
[27]	Dong J J, Tung Y H, Chen C C, et al. Discriminant analysis of the geomorphic characteristics and stability of landslide dams[J]. Geomorphology, 2009, 110(3): 162-171.
[28]	Dong J J, Tung Y H, Chen C C, et al. Logistic regression model for predicting the failure probability of a landslide dam[J]. Engineering Geology, 2011, 117(1/2): 52-61.
[29]	柴贺军, 刘汉超, 张倬元. 大型崩滑堵江事件及其环境效应研究综述[J]. 地质科技情报, 2000, 19(2): 87-90. doi: 10.3969/j.issn.1000-7849.2000.02.021 Chai H J, Liu H C, Zhang Z Y. The study on the environment and environmental effects of the collapse and closure of the Yangtze River[J]. Geological Science and Technology Information, 2000, 19(2): 87-90(in Chinese with English abstract). doi: 10.3969/j.issn.1000-7849.2000.02.021
[30]	黄健, 贺子城, 黄祥, 等. 基于地貌特征的滑坡堰塞坝形成敏感性研究[J]. 地质科技通报, 2021, 40(5): 253-262. doi: 10.19509/j.cnki.dzkq.2021.0040 Huang J, He Z C, Huang X, et al. Formation sensitivity of landslide dam based on geomorphic characteristics[J]. Bulletin of Geological Science and Technology, 2021, 40(5): 253-262(in Chinese with English abstract). doi: 10.19509/j.cnki.dzkq.2021.0040
[31]	单熠博, 陈生水, 钟启明. 堰塞体稳定性快速评价方法研究[J]. 岩石力学与工程学报, 2020, 39(9): 1847-1859. https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202009011.htm Shan Y B, Chen S S, Zhong Q M. A rapid evaluation method of landslide dam stability[J]. Chinese Journal of Rock Mechanics and Engineering, 2020, 39(9): 1847-1859(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-YSLX202009011.htm
[32]	罗宗伟. 抛掷爆破技术在马鞍石堰塞湖应急排险处置中的应用[J]. 水利水电技术, 2008(8): 31-32, 35. doi: 10.3969/j.issn.1000-0860.2008.08.010 Luo Z W. Application of casting blast to emergency treatment of risk-elimination of Maanshi Landslide Dam[J]. Water Resources and Hydropower Engineering, 2008(8): 31-32, 35(in Chinese with English abstract). doi: 10.3969/j.issn.1000-0860.2008.08.010
[33]	柴贺军, 刘汉超, 张倬元, 等. 天然土石坝稳定性初步研究[J]. 地质科技情报, 2001, 20(1): 77-81. doi: 10.3969/j.issn.1000-7849.2001.01.017 Chai H J, Liu H C, Zhang Z Y, et al. Preliminarily stability analysis of natural rock field dam resulting from damming landslide[J]. Geological Science and Technology Information, 2001, 20(1): 77-81(in Chinese with English abstract). doi: 10.3969/j.issn.1000-7849.2001.01.017
[34]	张宗亮, 程凯, 杨再宏, 等. 红石岩堰塞坝应急处置与整治利用关键技术[J]. 水电与抽水蓄能, 2020, 6(2): 1-10, 25. https://www.cnki.com.cn/Article/CJFDTOTAL-DBGC202002002.htm Zhang Z L, Cheng K, Yang Z H, et al. Key technology of emergency remedy and treatment for hongshiyan barriers lake dam[J]. Hydropower and Pumped Storage, 2020, 6(2): 1-10, 25(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-DBGC202002002.htm
[35]	刘建康, 程尊兰, 佘涛. 云南鲁甸红石岩堰塞湖溃坝风险及其影响[J]. 山地学报, 2016, 34(2): 208-215. https://www.cnki.com.cn/Article/CJFDTOTAL-SDYA201602013.htm Liu J K, Cheng Z L, She T. Assessment of dam failure and secondary hazards for Hongshiyan dammed lake caused by Ludian earthquake in Niulanjiang River[J]. Journal of Mountain Science, 2016, 34(2): 208-215(in Chinese with English abstract). https://www.cnki.com.cn/Article/CJFDTOTAL-SDYA201602013.htm