深基坑工程因其复杂环境与多变地质条件,支护结构变形引发的坍塌事故频发,传统监测方法(如测斜仪)受成本与布点密度限制,导致数据样本量小、时序稀疏,严重制约了基于深度学习的支护结构深层水平位移预测模型的泛化能力与预警精度。为突破小样本约束并提升预测性能,以郑州千玺广场深基坑工程为例,采用数据增强技术对支护结构变形监测数据集进行有效扩充,并系统评估了该技术对长短期记忆网络、门控循环单元、卷积神经网络及iTransformer等深度学习模型预测性能的影响。结果表明:数据增强能显著克服样本不足的限制,大幅提高各模型对基坑变形的预测精度;当样本扩增至约40倍时,卷积神经网络、长短期记忆网络与门控循环单元模型的性能指标提升率均超过95%,其综合表现优于iTransformer模型。研究可为基坑失稳精准预警提供参考。
Deep excavation engineering, due to its complex environment and variable geological conditions, frequently experiences collapse accidents caused by the deformation of supporting structures. Traditional monitoring methods (such as inclinometers) are limited by cost and layout density, resulting in small data sample sizes and sparse time series. Insufficient samples severely restrict the generalization capability and warning accuracy of deep learning-based prediction models for deep-seated horizontal displacement of supporting structures. To overcome the constraints of small samples and enhance prediction performance, this study takes the deep excavation project of Zhengzhou Qianxi Plaza as a case study. Data augmentation technology is applied to effectively expand the deformation monitoring dataset of the supporting structure, and its impact on the prediction performance of deep learning models—including Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), Convolutional Neural Network (CNN), and iTransformer—is systematically evaluated. The results show that:Data augmentation significantly overcomes the limitation of insufficient samples and greatly improves the prediction accuracy of each model for excavation deformation. When the sample size is expanded by approximately 40 times, the performance improvement rates of CNN, LSTM, and GRU models all exceed 95%, and their overall performance surpasses that of the iTransformer model. This research provides an effective new approach to overcoming the small-sample bottleneck of deep learning in excavation safety monitoring and achieving precise early warning of supporting structure deformation, holding significant engineering application value.
[1] 刘新荣,刘鹏,周小涵,等.基于变形影响的基坑邻近建筑保护措施研究[J].地下空间与工程学报,2024,20(1):263-275.
[2] 郑刚,石建成,程雪松,等. 天津某地铁车站深基坑突涌事故分析[J].建筑结构, 2024,54(21):132-137.
[3] 周越洲,杨慎银,王嵩,等. 深厚淤泥软土地区某基坑坍塌事故分析及处理[J].建筑结构, 2024,54(15):128-135.
[4] 程雪松,裴昊田,衣凡,等. 多道撑式深基坑垮塌事故连续破坏机理研究[J].土木工程学报,2024,57(3):102-109.
[5] 张旷成,李继民. 杭州地铁湘湖站“08.11.15”基坑坍塌事故分析[J].岩土工程学报,2010,32(增1):338-342.
[6] 陈绍清,熊思斯,何朝远,等. 地铁深基坑坍塌事故安全风险分析[J].安全与环境学报,2020,20(1):52-58.
[7] 夏元友,陈春舒,陈金培,等. 基于现场监测的深基坑施工动态风险评估[J].地下空间与工程学报, 2016,12(5):1378-1384.
[8] 崔高航,刘守花,王兆亮. 哈尔滨市某桩锚支护深基坑工程监测[J].建筑结构学报,2016,37(7):144-150.
[9] 冯利坡,郑永来,赵民,等. 信息化监测在深基坑工程事故中的预警作用[J].地下空间与工程学报,2009, 5(增2):1643-1645,1652.
[10] 杨帅帅,白羽,缪易辰,等. 深基坑开挖变形研究进展[J].工业安全与环保, 2023,49(1):37-42.
[11] Sepp H, Jürgen S. Long Short-Term Memory[J]. Neural Computation, 1997, 9 (8):1735-1780.
[12] Zhang Q, Ma Y, Zhang B, et al. Time Series Prediction on Settlement of Metro Tunnels Adjacent to Deep Foundation Pit by Clustering Monitoring Data[J]. KSCE Journal of Civil Engineering, 2023, 27:2180-2190.
[13] Côté P, Nikanjam A, Ahmed N, et al.Data cleaning and machine learning:a systematic literature review[J]. Automated Software Engineering, 2024, 31(2):1-75.
[14] 徐长节,李欣雨. 基于人工神经网络的深基坑支护结构侧移预测[J].上海交通大学学报,2024,58(11):1735-1744.
[15] Li T, Shu J, Wang Y.Deformation prediction of underground engineering support structures via the ST-CNN-LSTM model[J]. Journal of Civil Structural Health Monitoring, 2025, 15:1901-1919
[16] Ma G, Zang X, Chen S, et al. The forecasting of surface displacement for tunnel slopes utilizing the WD-IPSO-GRU model[J].Scientific Reports, 2024, 14(1):1-21.
[17] 黄雨,刘侃侃,程天笑,等. 基于时间融合Transformer的港池基坑开挖诱发形变的智能预测模型[J].应用基础与工程科学学报, 2025,33(5):1287-1296.
[18] Guo J, Jia N, Bai J. Transformer based on channel-spatial attention for accurate classification of scenes in remote sensing image[J]. Scitific Reports, 2022, 12:15473.
[19] 郑远攀,李广阳,李晔. 深度学习在图像识别中的应用研究综述[J].计算机工程与应用,2019,55(12):20-36.
[20] 高友文,周本君,胡晓飞. 基于数据增强的卷积神经网络图像识别研究[J].计算机技术与发展,2018,28(8):62-65.
[21] 徐勇. 基于深层神经网络的语音增强方法研究[D].合肥:中国科学技术大学,2015. (Xu Yong. Research on Deep Neural Network based Speech Enhancement[D]. Hefei:University of Science and Technology of China, 2015. (in Chinese))
[22] Steven Y F, Varun G, Jason W, et al.A Survey of Data Augmentation Approaches for NLP[A]//In Findings of the Association for Computational Linguistics:ACL-IJCNLP[C], 2021:968-988.
[23] Qiu H, Chen X, Feng P, et al.Advancing predictive accuracy of shallow landslide using strategic data augmentation[J]. Journal of Rock Mechanics and Geotechnical Engineering, 2025, 17(7):4273-4287.
[24] Wen Q, Sun L, Yang F, et al. Time Series Data Augmentation for Deep Learning:A Survey[A]//Proceedings of the Thirtieth International Joint Conference on Artificial Intelligence Survey Track[C]. 2021. 4653-4660.
[25] Iwana B K, Uchida S. An empirical survey of data augmentation for time series classification with neural networks[J]. PLoS ONE, 2021, 16(7):e0254841.
[26] Liu Y, Hu T, Zhang H, et al. iTransformer:Inverted Transformers Are Effective for Time Series Forecasting[A]//12th International Conference on Learning Representations, ICLR[C].2024.
[27] Xu Y, Hu C, Wu Q, et al. Research on particle swarm optimization in LSTM neural networks for rainfall-runoff simulation[J]. Journal of Hydrology, 2022, 608:127553.
[28] Karasu S, Altan A. Crude oil time series prediction model based on LSTM network with chaotic Henry gas solubility optimization[J]. Energy, 2022, 242, 122964.
[29] Le Cun Y, Bengio Y, Hinton G. Deep learning[J]. Nature, 2015, 521:436-444.
[30] Cha Y, Choi W, Büyüköztürk O. Deep learning-based crack damage detection using convolutional neural networks[J]. Computer-Aided Civil and Infrastructure Engineering, 2017,32:361-378.
[31] Bao X, Fan T, Shi C. et al. One-dimensional convolutional neural network for damage detection of jacket-type offshore platforms[J]. Ocean Engineering, 2021, 219:108293.
[32] Simo-Serra E, Iizuka S, Ishikawa H. Mastering sketching:adversarial augmentation for structured prediction[A]//ACM Transactions on Graphics (TOG)[C], 2018, 37:1-13.