Intelligent Rock Mass Quality Evaluation Considering Geological Heterogeneity of Tunnel Face Using Drilling Parameters

Rock mass quality evaluation dictates excavation method and support structure design in tunnel engineering. This study presented a quantitative method to characterize tunnel face geological heterogeneity through drilling parameter images. Image recognition techniques subsequently facilitated intelligent rock mass quality evaluation. The methodology began with data processing on a large dataset of drilling parameters. A spatial mesh enabled data interpolation and normalization to generate a normalized grid using Python (version 3.7) toolkits. Sample images of borehole drilling parameters were generated for each tunnel face using the CMYK mode. Subsequently, an intelligent rock mass evaluation model was constructed via transfer learning using the Inception-V3 convolutional neural network. The trained model achieved an accuracy of 96%. Conventional approaches average drilling parameters across all boreholes at a tunnel face. In contrast, the proposed method incorporated spatial geological heterogeneity. This approach significantly enhanced intelligent rock mass evaluation. The findings reduce the time and financial cost of geotechnical investigations during tunnel construction.