水封洞库Q分值围岩分级深度学习优化
Deep Learning Optimization of Q-Value Method of Water-Sealed Cave Library
DOI: 10.12677/ME.2023.112034, PDF,   
作者: 刘珠兰：青岛理工大学理学院，山东 青岛；韩立亮, 孙 健：中石油华东设计院有限公司，山东 青岛
关键词: Q值法；卷积神经网络；地下水封洞库；图像处理技术；Q-Value Method； CNN； Underground Oil Storage in Rock Caverns； Image Processing Technology
摘要: 水封洞库现场实地调研发现Q值法受施工人员、环境、设备等因素影响较大，分级工作流于形式，论文利用图像处理手段对其进行优化提高分级结果准确性。首先，对水封洞库现场地质素描围岩分级结果整理，将围岩等级和岩体发育程度分别划分为四梯度，节理面强度和赋存条件分别划分为三梯度；然后，将采集到的围岩图像批量裁剪与筛选，并借用数据增强技术完成现场图像分类数据构建；最后将训练集和验证集输入到分类网络(EfficentNet_B3)中进行训练，训练结束后再对模型进行验证。
Abstract: The field investigation of the water-sealed cave warehouse found that the Q-value method was greatly affected by the construction personnel, environment, equipment and other factors, and the grading workflow was in form. The paper used image processing to optimize it to improve the ac-curacy of the grading results. Firstly, the grading results of the surrounding rock of the geological sketch at the site of the water-sealed cave reservoir were sorted out, and the surrounding rock grade and rock mass development degree were divided into four gradients, and the joint surface strength and occurrence conditions were divided into three gradients. Then, the collected sur-rounding rock images are clipped and screened in batches, and the data enhancement technology is used to complete the construction of on-site image classification data. Finally, the training set and validation set are input to the classification network (EfficentNet_B3) for training, and the model is verified after training.
文章引用：刘珠兰, 韩立亮, 孙健. 水封洞库Q分值围岩分级深度学习优化[J]. 矿山工程, 2023, 11(2): 276-285. https://doi.org/10.12677/ME.2023.112034

参考文献

[1] Barton, N.R., Lien, R. and Lunde, J. (1974) Engineering Classification of Rock Masses for the Design of Rock Support. Rock Me-chanics, 6, 189-236. [Google Scholar] [CrossRef] 
[2] 马云飞. 隧道围岩Q值分级与支护设计优化[J]. 交通科技, 2013(2): 83-84.
[3] 康小兵, 许模, 陈旭. 岩体质量Q系统分类法及其应用[J]. 中国地质灾害与防治学报, 2008, 19(4): 91-95.
[4] 刘现春, 石豫川, 童建刚. Q系统在西南某水电TBM施工隧洞围岩质量分类中的应用[J]. 水资源与水工程学报, 2011, 22(5): 167-170.
[5] 肖春华. 岩体质量指标Q分类法在汕头LPG工程中的应用[J]. 隧道建设, 2002(4): 4-9.
[6] Xu, D., Huang, X., Li, S., et al. (2022) Predicting the Excavation Damaged Zone within Brittle Surrounding Rock Masses of Deep Under-ground Caverns Using a Comprehensive Approach Integrating in Situ Measurements and Numerical Analysis. Geoscience Frontiers, 13, Article ID: 101273. [Google Scholar] [CrossRef] 
[7] 史永跃, 李苍松, 李春林, 等. 某地下油库蚀变岩工程特性及岩体质量评价技术研究[J]. 现代隧道技术, 2019, 56(Z1): 150-158.
[8] 冷彪. 数字图像处理在隧道掌子面图表像中的研究与实现[D]: [硕士学位论文]. 成都: 西南交通大学, 2005.
[9] 李汪石. 公路隧道围岩云分级系统的开发与应用[D]: [硕士学位论文]. 长沙: 长沙理工大学, 2017.
[10] Dongho, K., Sukhpreet, S.B., Dharshan, L.G. and Young-Jin, C. (2020) Hybrid Pixel-Level Concrete Crack Segmentation and Quantification Across Complex Backgrounds Using Deep Learning. Automation in Construction, 118, Article ID: 103291. [Google Scholar] [CrossRef] 
[11] Jia, A.K., Xue, X.L., Wang, Y., Luo, X.W. and Xue, W.R. (2020) An Integrated Approach to Automatic Pixel-Level Crack Detection and Quantification of Asphalt Pavement. Automation in Construction, 114, Article ID: 103176. [Google Scholar] [CrossRef] 
[12] Amir, R., Radhakrishna, A., Michele, G. and Katrin, B. (2020) Comparison of Crack Segmentation Using Digital Image Correlation Measurements and Deep Learning. Construction and Building Materials, 261, Article ID: 120474. [Google Scholar] [CrossRef] 
[13] 朱鑫. 隧道结构表面病害特征快速检测研究[D]: [硕士学位论文]. 成都: 西南交通大学, 2014.
[14] 查焕奕. 基于ResNet模型的隧道围岩云分级系统开发[D]: [硕士学位论文]. 长沙: 长沙理工大学, 2018.
[15] 柳厚祥, 李汪石, 查焕奕, 蒋武军, 许腾. 基于深度学习技术的公路隧道围岩分级方法[J]. 岩土工程学报, 2018, 40(10): 1809-1817.
[16] 李赤谋, 吕明, 袁青, 陈宇佳, 王树英. 基于三维重建与Unet神经网络的隧道掌子面围岩快速分级技术[J]. 隧道建设(中英文), 2022, 42(1): 33-40.
[17] He, K., Zhang, X., Ren, S. and Sun, J. (2016) Deep Residual Learning for Image Recognition. Proceedings of 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, 27-30 June 2016, 770-778. [Google Scholar] [CrossRef] 
[18] Tan, M. and Le, Q. (2019) EfficientNet: Rethinking Model Scaling for Convolu-tional Neural Networks. Proceedings of 36th International Conference on Machine Learning, Long Beach, 9-15 June 2019, 6105-6114.

为你推荐