In order to overcome the strong dependence of traditional acoustic wave testing methods on testing equipment and site conditions in tunnel rock mass integrity evaluation, and to achieve rapid and convenient prediction of the rock mass integrity coefficient (Kv), an integrated rock mass integrity assessment method based on deep learning is proposed. Taking surrounding rock fracture images as input, a fracture image dataset is constructed, and a ConvNeXt-based region convolutional neural network model is introduced to realize automatic fracture identification and semantic segmentation. Subsequently, image processing techniques are employed to quantitatively analyze the identified fractures, from which seven geometric features, including fracture length, density, number of intersections, and roughness are extracted. Furthermore, a genetic programming algorithm is utilized to establish a nonlinear mapping relationship between fracture feature parameters and Kv. The results indicate that the proposed model achieves a recall rate of 91.83% in fracture identification. Cross-validation based on the genetic programming model shows that the root mean square error of Kv prediction is 0.026 3±0.009 4, and the mean absolute percentage error is 2.57%±0.46%. The proposed method effectively enables Kv prediction based on fracture image features, providing an alternative approach to rock mass integrity evaluation that does not rely on acoustic wave testing and demonstrates promising engineering application potential.
CAIJianhua, ZHANGHaichao, WANGXu. Research of the arrangement method for quick testing rock-mass integrity index on tunnel face[J]. Modern Tunnelling Technology, 2019, 56(2): 164-169.
YINMinglun, ZHANGJinxun, JIANGYusheng, et al. Study of correction of the structural plane category based on the rock mass integrity coefficient characterized by the volumetric joint count [J]. Rock and Soil Mechanics, 2021, 42(4): 1133-1140.
[6]
HASANM, SHANGY J, YIX T, et al. Determination of rock mass integrity coefficient using a non-invasive geophysical approach [J]. Journal of Rock Mechanics and Geotechnical Engineering, 2023, 15(6): 1426-1440.
[7]
WANGJ C, HANZ Q, WANGC, et al. Rock mass integrity evaluation method of roadway surrounding rock mass based on borehole image acoustic radar data [J]. Scientific Reports, 2025, 15: 22050.
FUJinyang, LIANGXiangrong, WANGYu, et al. Intelligent identification of discontinuity in surrounding rock of tunnels based on three-dimensional superpixel segmentation [J]. China Journal of Highway and Transport, 2023, 36(11): 289-301.
[11]
ZHAOG B, RUANM D, TUNCAYD, et al. Investigation on capturing bedding planes in laminated shale through advanced physics-informed image processing for multiscale geomechanical simulation [J]. Engineering Geology, 2025, 347: 107929.
WANGJunxiang, ZENGXiangsen, XUChenhui, et al. Study on quantitative identification method of rock fracture based on image processing technology[J]. Chinese Journal of Underground Space and Engineering, 2022, 18(2): 446-457.
[14]
WUX, WANGF Y, CHENJ P, et al. Automatic extraction of tunnel surrounding rock fractures based on FPD-net and integrated post-processing technology[J]. Tunnelling and Underground Space Technology, 2025, 166: 106993.
WANGChunyan, WANGKangle, JIANGYong, et al. Crack extraction of concrete pavement based on U2-Net and dynamic index-based rotational convolution[J]. Journal of Liaoning Technical University (Natural Science), 2025, 44(6): 746-752.
[17]
CHAIYASARNK, BUATIKA, MOHAMADH, et al. Integrated pixel-level CNN-FCN crack detection via photogrammetric 3D texture mapping of concrete structures[J]. Automation in Construction, 2022, 140: 104388.
[18]
De CARVALHOO L F, De CARVALHO JÚNIORO A, De ALBUQUERQUEA O, et al. Instance segmentation for large, multi-channel remote sensing imagery using mask-RCNN and a mosaicking approach[J]. Remote Sensing, 2021, 13(1): 39.
[19]
LIQ L, XUX Y, GUANJ J, et al. The improvement of faster-RCNN crack recognition model and parameters based on attention mechanism [J]. Symmetry, 2024, 16: 1027.
[20]
LYUH T, QIUF, ANL, et al. Deer survey from drone thermal imagery using enhanced Faster R-CNN based on ResNets and FPN [J]. Ecological Informatics, 2024, 79: 102383.
YANGYuxuan, ZHONGBaorong. Method of lightweight rock debris lithology identification based on improved ConvNeXt model [J]. Modern Electronics Technique, 2024, 47(24): 131-136.
[23]
WANGD F, HUH F, CHEND H. Transformer with sparse self-attention mechanism for image captioning [J]. Electronics Letters, 2020, 56(15): 764-766.
[24]
RUSSELLB C, TORRALBAA, MURPHYP K P, et al. LabelMe: a database and web-based tool for image annotation [J]. International Journal of Computer Vision, 2008, 77(1/2/3): 157-173.
ZHOUXin, XUPeizhe, LIKun, et al. Optimization of YOLOv8 flame target detection algorithm based on improved loss function and attention mechanism[J]. Ship & Ocean Engineering, 2025, 54(2): 19-25.
[27]
ZHAOY, HEX, JIANGL S, et al. Influence analysis of complex crack geometric parameters on mechanical properties of soft rock [J]. International Journal of Coal Science & Technology, 2023, 10: 78.
[28]
WANGS W, XUJ B, WUX, et al. Concrete crack recognition and geometric parameter evaluation based on deep learning [J]. Advances in Engineering Software, 2025, 199: 103800.
[29]
QINY, XUN X, HANJ, et al. Experimental study on the effects of geometric parameters of filled fractures on the mechanical properties and crack propagation mechanisms of rock masses [J]. Rock Mechanics and Rock Engineering, 2023, 56(4): 2697-2716.
WUShunchuan, SUNBeibei, YUYisong, et al. Multi-scale image recognition method of rock mass fracture traces considering geometric features [J]. Journal of Huazhong University of Science and Technology (Natural Science Edition), 2024, 52(9): 101-109, 117.
XIEYuan, GAOWei, WANGYiwei, et al. Life prediction of RC underground structure by sulfate attack based on harris hawks optimizing genetic programming [J]. China Civil Engineering Journal, 2022, 55(4): 33-41.