基于LSTM-熵权TOPSIS的双排抗滑桩地震损伤动态评价

Dynamic evaluation of seismic damage of double-row anti-slide piles based on LSTM-entropy weighted TOPSIS

摘要

地震频发区边坡地震稳定性问题已成为工程自然灾害领域的研究热点。抗滑桩作为边坡抗震加固的核心构件，其地震损伤演化机制研究对保障工程安全具有重要意义。针对当前抗滑桩地震响应损伤预测评价研究不足的问题，本文通过振动台模型试验构建双排抗滑桩-边坡体系振动响应测试系统，同步采集不同峰值地面加速度(peak ground acceleration，PGA)地震波作用下桩体加速度及动土压力数据。基于长短期记忆网络(long short-term memory，LSTM)建立地震损伤位移预测模型，结合熵权法-逼近理想解排序法(entropy weight method-technique for order preference by similarity to ideal solution，EW-TOPSIS)构建多指标耦合损伤评价体系。研究结果表明，LSTM模型通过捕捉桩-土体系强非线性特征，显著提升了位移预测精度，验证了模型的可靠性。熵权法权重分析显示，位移指标占损伤评估的主导地位，其中水平向(X)地震波主要触发桩基部位移，纵向(Z)波则显著影响桩顶位移，随着峰值地面加速度增大，位移指标权重呈非线性增长，揭示桩-土体系在强震下由局部变形向整体失稳转变的损伤演化机制。熵权-TOPSIS综合评价模型证实双排桩损伤呈现显著空间异质性，山侧桩损伤程度较河侧桩提升47.2%，后排桩损伤较前排桩降低26.7%，并发现0.2 g PGA为损伤模式转换临界点。以桩身位移和动土压力为指标，建立了抗滑桩地震响应损伤预测评估模型，进而为抗震结构设计优化、结构健康监测以及边坡治理提供支撑。

Abstract

The seismic stability of slopes in earthquake-prone areas has become a hot research topic in the field of engineering natural disasters. As the core component for seismic reinforcement of slopes, the study of the seismic damage evolution mechanism of anti-sliding piles is of great significance for ensuring engineering safety. In response to the current lack of research on the seismic response damage prediction and evaluation of anti-sliding piles, this paper constructs a vibration response testing system for a double-row anti-sliding pile-slope system using shaking table model tests, simultaneously collecting data on pile acceleration and dynamic soil pressure under seismic waves with different peak ground accelerations (PGA). A displacement prediction model for seismic damage is developed using the long short-term memory (LSTM) network, and a multi-indicator coupled damage evaluation framework is constructed by integrating the entropy weight method and technique for order preference by similarity to ideal solution (EW-TOPSIS). The results show that the LSTM model significantly improves the accuracy of displacement prediction by capturing the strong nonlinear characteristics of the pile-soil system, validating the reliability of the model. The entropy weight analysis indicates that displacement metrics dominate the damage assessment. The horizontal (X) seismic waves mainly trigger the displacement at the pile base, while the vertical (Z) waves significantly affect the pile top displacement. As the PGA increases, the weight of the displacement index increases non-linearly, revealing the damage evolution mechanism of the pile-soil system, from local deformation to overall instability under strong seismic conditions. The entropy-weight-TOPSIS integrated evaluation model confirms that the damage of the double-row piles exhibits significant spatial heterogeneity, with the mountain-side piles having 47.2% more damage than the river-side piles, and the damage of the rear piles being 26.7% lower than the front piles. Furthermore, 0.2 g PGA is identified as the critical point for damage mode conversion. Using pile displacement and dynamic soil pressure as indicators, a seismic response damage prediction and evaluation model for anti-sliding piles is established, providing support for seismic structural design optimization, structural health monitoring, and slope management.

关键词

双排抗滑桩 / 振动台试验 / 长短期记忆网络 / 熵权-逼近理想解排序法 / 损伤评价

Key words

double-row anti-slide pile / shaking table test / long short-term memory / EW-TOPSIS / damage assessment

参考文献

[1] 中华人民共和国应急管理部. 应急管理部发布2022年全国十大自然灾害[EB/OL]. (2023-01-12) [2025-08-21]. https://www.mem.gov.cn/xw/yjglbgzdt/202301/t20230112_440396.shtml

[2] Ministry of Emergency Management of the People’s Republic of China. Basic information on natural disasters in China in 2022[EB/OL]. (2023-01-13). [2025-08-21]. https://www.mem.gov.cn/xw/yjglbgzdt/202301/t20230112_440396.shtml. (in Chinese)

[3] Gautam P, Kubota T, Aditian A. Evaluating underlying causative factors for earthquake-induced landslides and landslide susceptibility mapping in Upper Indrawati Watershed, Nepal[J]. Geoenvironmental Disasters, 2021, 8(1): 30.

[4] 朱丹, 蒋关鲁, 陈虹羽, 等 . 单、双排桩支护路堑边坡动力响应特性的振动台试验研究[J]. 岩土力学, 2024, 45(6): 1763-1777.

[5] Zhu Dan, Jiang Guanlu, Chen Hongyu, et al. Shaking table experimental study on the dynamic response characteristics of single and double-row pile-supported road graben slopes[J]. Rock and Soil Mechanics, 2024, 45(6): 1763-1777. (in Chinese)

[6] Zhu Baolong, Ren Long, Feng Kang, et al. Investigation of the dynamic response of h-type anti-slide pile based on shaking table test[J]. Soil Dynamics and Earthquake Engineering, 2024, 182: 108736.

[7] 孔思宇. 基于计算机视觉与数据驱动的结构裂缝检测与损伤评估[D]. 北京: 清华大学, 2023.

[8] Kong Siyu. Crack detection and structural damage assessment based on computer vision and data-driven methods[D]. Beijing: Tsinghua University, 2023. (in Chinese)

[9] 毛晨曦, 郭永超, 张昊宇, 等 . 基于卷积神经网络的RC框架通信机楼震后损伤评定方法[J]. 自然灾害学报, 2024, 33(5): 157-167.

[10] Mao Chenxi, Guo Yongchao, Zhang Haoyu, et al. Post-earthquake damage assessment for RC frame communication buildings based on convolutional neural network[J]. Journal of Natural Disasters, 2024, 33(5): 157-167. (in Chinese)

[11] 唐华都. 基于深度学习的建筑物损伤程度评估方法[D]. 秦皇岛: 燕山大学, 2024.

[12] Tang Huadu. A deep learning-based method for assessing the damage of buildings[D]. Qinhuangdao: Yanshan University, 2024. (in Chinese)

[13] Yao Mingxia, Tian Dongdong, Hu Guohui, et al. Research on landslide displacement prediction based on transformer[J]. ce/papers, 2025, 8(2): 613-619.

[14] 王沛宇, 吕润田, 缪杰蔚, 等 . 强震区抗滑桩加固碎石土边坡动力响应特征与桩身损伤识别: 以川青铁路九寨沟双线大桥附近工点为例[J]. 科学技术与工程, 2024, 24(25): 10910-10920.

[15] Wang Peiyu, Lv Runtian, Miao Jiewei, et al. Dynamic response characteristics and damage identification of pile body of gravel soil slope strengthened by anti-slide pile in strong earthquake area: take Jiuzhaigou double-line bridge of Sichuan-Qing railway as an example[J]. Science Technology and Engineering, 2024, 24(25): 10910-10920. (in Chinese)

[16] 田鸿程. 抗滑桩损伤识别及评估方法的振动台试验研究[D]. 成都: 西南交通大学, 2020.

[17] Tian Hongcheng. Shaking table experimental research on damage identification and evaluation method of antislide pile[D]. Chengdu: Southwest Jiaotong University, 2020. (in Chinese)

[18] Gupta P, Mehn diratta S. Exploring the efficacy of slope stabilization using piles: a comprehensive review[J]. Indian Geotechnical Journal, 2025, 55(5): 3442-3459.

[19] 苏杭. 基于模糊数学与层次分析法的抗滑桩震害评估[J]. 甘肃水利水电技术, 2019, 55(12): 17-21.

[20] Su Hang. Seismic damage assessment of antisliding piles based on fuzzy mathematics and analytic hierarchy process[J]. Gansu Water Resources and Hydropower Technology, 2019, 55(12): 17-21. (in Chinese)

[21] 杨羽, 韦洪, 余浩, 等 . 地震作用下多锚点圆形抗滑桩动力响应特性研究[J]. 自然灾害学报, 2024, 33(5): 61-72.

[22] Yang Yu, Wei Hong, Yu Hao, et al. Research on dynamic response characteristics of multi-anchor circular anti-slide pile under earthquake[J]. Journal of Natural Disasters, 2024, 33(5): 61-72. (in Chinese)

[23] Hu Bin, Han Yaohui, Zhang Wenhui, et al. Correction: a prediction approach to COVID-19 time series with LSTM integrated attention mechanism and transfer learning[J]. BMC Medical Research Methodology, 2025, 25(1): 29.

[24] 王守华, 王睿菘, 孙希延, 等 . 基于CF-CNN-LSTM模型的滑坡易发性评价[J]. 自然灾害学报, 2024, 33(5): 84-95.

[25] Wang Shouhua, Wang Ruisong, Sun Xiyan, et al. Landslide susceptibility evaluation based on CF-CNN-LSTM model[J]. Journal of Natural Disasters, 2024, 33(5): 84-95. (in Chinese)

[26] Xu Biao, Rang Gejie, Xie Ruijie, et al. A prediction approach based on long short-term memory networks for dynamic multiobjective optimization[J]. Expert Systems with Applications, 2025, 283: 127792.

[27] 谢豪, 王靖宇. 基于LSTM神经网络的隧道衬砌病害识别方法研究[J]. 地球物理学进展, 2025, 40(5): 2227-2236.

[28] Xie Hao, Wang Jingyu. Research on tunnel lining disease identification method based on LSTM neural network[J]. Progress in Geophysics, 2025, 40(5): 2227-2236. (in Chinese)

[29] Ai Xueshan, Yu Yangxin, Liang Zhiming, et al. Multilayer entropy-weighted TOPSIS method and its decision-making in ecological operation during the subsidence period of the Three Gorges Reservoir[J]. Scientific Reports, 2025, 15: 2954.

[30] 巴婉茹, 宫阿都, 张树宇, 等 . 基于熵权TOPSIS模型的古建筑地震灾害风险评估[J]. 自然灾害学报, 2024, 33(4): 1-10.

[31] Ba Wanru, Gong Adu, Zhang Shuyu, et al. Earthquake hazard assessment of ancient buildings based on entropy weight TOPSIS model[J]. Journal of Natural Disasters, 2024, 33(4): 1-10. (in Chinese)

[32] Gao Tiaofang, Chen Liang, Zhang Xinghui, et al. Credibility assessment for digital twins in vehicle-in-the-loop test based on information entropy[J]. Sensors, 2025, 25(5): 1372.

[33] Park H, Kim K, Hyung J, et al. Decision-making for the hazard ranking of water distribution networks using the TOPSIS method[J]. Water Supply, 2023, 23(2): 715-726.

[34] 孙同刚. 泰禾集团财务风险评价及应对研究[D]. 青岛: 青岛大学, 2024.

[35] Sun Tonggang. Research on financial risk assessment and response of taihe group[D]. Qingdao: Qingdao University, 2024. (in Chinese)

[36] 李平, 刘应慈, 周楷, 等 . 振动台模型试验相似设计综述[J]. 防灾科技学院学报, 2020, 22(4): 29-35.

[37] Li Ping, Liu Yingci, Zhou Kai, et al. Review of similar design of shaking table model tests[J]. Journal of Institute of Disaster Prevention, 2020, 22(4): 29-35. (in Chinese)

[38] 艾挥, 吴红刚, 冯文强, 等 . 冠梁约束型抗滑桩与普通抗滑桩的抗震性能对比试验研究[J]. 防灾减灾工程学报, 2017, 37(2): 194-200.

[39] Ai Hui, Wu Honggang, Feng Wenqiang, et al. Research on seismic capacity correlative experiment of top beam binding anti-slide pile and common anti-slide pile[J]. Journal of Disaster Prevention and Mitigation Engineering, 2017, 37(2): 194-200. (in Chinese)

[40] Wu Honggang, Pai Lifang. Shaking table test for reinforcement of soil slope with multiple sliding surfaces by reinforced double-row anti-slide piles[J]. Journal of Mountain Science, 2022, 19(5): 1419-1436.

[41] Yang Fei, Xin Chunlei, Wang Zhao, et al. Investigating deformation and failure mechanisms of discontinuous anti-dip bench rock slopes through shaking table tests and numerical simulations[J]. Rock Mechanics and Rock Engineering, 2025, 58(5): 5329-5361.

[42] Zou Zuyin, Lei Da, Jiang Guanlu, et al. Experimental study of bridge foundation reinforced with front and back rows of anti-slide piles on gravel soil slope under El Centro waves[J]. Applied Sciences, 2020, 10(9): 3108.

[43] Jugnesh K. Data warehouse and data mining: concepts, techniques and real life applications (English Edition)[M]. London: BPB Publishers, 2024.

[44] 缪惠全. 加速度基线漂移时域处理方法的对比研究[J]. 地震工程与工程振动, 2022, 42(2): 135-150.

[45] Miao Huiquan. Comparative study of time-domain processing methods of acceleration baseline drift[J]. Earthquake Engineering and Engineering Dynamics, 2022, 42(2): 135-150. (in Chinese)

[46] 李海霞, 宋丹蕾, 孔佳宁, 等 . 传统机器学习模型的超参数优化技术评估[J]. 计算机科学, 2024, 51(8): 242-255.

[47] Li Haixia, Song Danlei, Kong Jianing, et al. Evaluation of hyperparameter optimization techniques for traditional machine learning models[J]. Computer Science, 2024, 51(8): 242-255. (in Chinese)

[48] Hodson T O. Root-mean-square error (RMSE) or mean absolute error (MAE): when to use them or not[J]. Geoscientific Model Development, 2022, 15(14): 5481-5487.

基金资助

中国中铁股份有限公司科技研究开发计划项目(2022-重大专项-07)

甘肃省重点研发计划-工业类项目(23YFGA0028)

重庆市交通科技项目(CQJT2022ZC20)

国家铁路局课题(KF2026-036)