盾构运动轨迹偏移会造成管片错位及地面沉降等危害,本文提出了一种用于盾构掘进过程中轨迹偏差的实时预测深度学习模型。该模型将时域卷积网络(TCN)和控制门单元(GRU)结合起来,并引入注意力机制。首先利用小波变换去噪(WT)对盾构掘进过程中收集数据的噪声进行去除,通过TCN算法对输入的时间序列进行局部特征的提取,再采用GRU算法提取时间序列数据的长期依赖性特征,最后引入注意力机制,进一步增强模型对输入数据中重要信息的关注程度。以福州滨海快线第三标段为例,对本混合模型进行了验证,并与其他三种效果较好的深度学习模型进行比较。结果表明,该模型对盾构运动轨迹的预测精度高于其他模型,该预测框架为盾构掘进过程中盾构移动轨迹的实时预测提供了一种有前景的解决方案。
Shield tunnelling trajectory deviation will cause segment dislocation ground subsidence and other hazards. A deep learning model for real-time prediction of trajectory deviation during shield tunneling is proposed. The model combines a Temporal Convolutional Network (TCN) and Gated Gate Unit (GRU) and introduces an attention mechanism. First, wavelet transform denoising (WT) is used to remove the noise of the data collected during the shield excavation process, the local features of the input time series are extracted by the TCN algorithm, and then the GRU algorithm is used to extract the long-term dependence features of the time series data, Finally, the attention mechanism is introduced to further enhance the model's attention to important information in the input data. Taking the third section of the Fuzhou Binhai Express Line as an example, the hybrid model was verified and compared with other three deep learning models with better effects. The results show that the prediction accuracy of the model for the movement trajectory of the shield is higher than that of other models, and this prediction framework provides a promising solution for the real-time prediction of the movement trajectory of the shield during the excavation of the shield.
[1] Liu Q S, Huang X, Gong Q M, et al. Application and development of hard rock TBM and its prospect in China[J]. Tunneling and Underground Space Technology, 2016, 57: 33-46.
[2] Gong Q M, Yin L J, Ma H S, et al. TBM tunnelling under adverse geological conditions: an overview[J]. Tunnelling and Underground Space Technology, 2016, 57: 4-17.
[3] Lin S S, Shen S L, Zhou A, et al. Risk assessment and management of excavation system based on fuzzy set theory and machine learning methods[J]. Automation in Construction, 2021, 122: 103490.
[4] Wang L T, Yang X, Gong G F, et al. Pose and trajectory control of shield tunneling machine in complicated stratum[J]. Automation in Construction, 2018, 93:192-199.
[5] 胡长明, 李靓, 梅源, 等.盾构竖向姿态预测模型参数的全局敏感性分析[J]. 现代隧道技术, 2021, 58(2): 127-134.
[6] 丁智,董毓庆,张霄,等.盾构姿态变化对管片影响与控制研究及展望[J]. 科学技术与工程, 2021, 21(21): 8745-8756.
[7] 周奇才, 陈俊儒, 何自强, 等.盾构智能化姿态控制器的设计[J]. 同济大学学报(自然科学版), 2008, 36 (1): 76-80.
[8] 郭正刚. 基于机器学习的盾构姿态调整决策方法研究[D]. 大连: 大连理工大学, 2013.
[9] Zhang Z H, Ma L H. Attitude correction system and cooperative control of tunnel boring machine[J]. International Journal of Pattern Recognition and Artificial Intelligence, 2018, 32(11): 1859018.
[10] Yue M, Sun W, Hu P. Dynamic coordinated control of attitude correction for the shield tunneling based on load observer[J]. Automation in Construction, 2012, 24(7): 24-29.
[11] Zhou C, Xu H C, Ding L Y, et al. Dynamic prediction for attitude and position in shield tunneling: a deep learning method[J]. Automation in Construction, 2019, 105: 102840.
[12] Ghimire S, Deo R C, Raj N, et al. Deep solar radiation forecasting with convolutional neural network and long short-term memory network algorithms[J]. Applied Energy, 2019, 253: 113541.
[13] Bardhan A, Kardani N, GuhaRay A, et al. Hybrid ensemble soft computing approach for predicting penetration rate of tunnel boring machine in a rock environment[J]. Journal of Rock Mechanics and Geotechnical Engineering, 2021,13(6):1398-1412.
[14] Elbaz K, Shen S L, Zhou A N. Optimization of EPB shield performance with adaptive neuro-fuzzy inference system and genetic algorithm[J]. Applied Sciences, 2019, 9(4): 780.
[15] 吴惠明, 常佳奇, 李刚,等.基于支持向量机的盾构掘进姿态预测与施工参数优化方法[J]. 隧道建设(中英文), 2021, 41(增1): 11-18.
[16] Zhang N, Zhang N, Zheng Q, et al. Real-time prediction of shield moving trajectory during tunneling using GRU deep neural network[J]. Acta Geotechnica, 2022, 17: 1167-1182.
[17] Shen S L, Elbaz K, Shaban W M, et al. Real-time prediction of shield moving trajectory during tunneling[J]. Acta Geotechnica, 2022, 17: 1533-1549.
[18] Kong X X, Ling X Z, Tang L,et al. Random forest-based predictors for driving forces of earth pressure balance (EPB) shield tunnel boring machine (TBM)[J]. Tunnelling and Underground Space Technology, 2022, 122: 104373.
[19] Silverman B W, Vassilicos J C, Ramsey J B. The contribution of wavelets to the analysis of economic and financial data[J]. Philosophical Transactions of the Royal Society of London, 1999, 357: 2593-2606.
[20] Bai S, Kolter J Z, Koltun V. An empirical evaluation of generic convolutional and recurrent networks for sequence modeling. [J/OL]. https://arxiv.org/abs/1803.01271.
[21] He K, Zhang X, Ren S, et al. Deep residual learning for image recognition[A] // Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition[C].IEEE Open Journal on Immersive Dispalys, 2016: 770-778.
[22] Cho K, Merriёnboer B V, Gulcehre C. Learning phrase representations using RNN encoder-decoder for statistical machine translation [A] // Moschitti A, Ed. Proceeding of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP) [C]. Doha: Association for Computational Linguistics, 2014: 1724-1734.
[23] Ashish V, Noam S, Niki P, et al. Attentionis all you need [J/OL]. https://arxiv.org/abs/1706.03762.
