A Study on Neural Network Evaluation Model of Blasting Effect in Mountain Tunnel and Decision-making Method for Blasting Parameter Optimization
(1. School of Civil Engineering, Central South University, Changsha 410075; 2. Guizhou Road & Bridge Group Co., Ltd.,Guiyang 550000; 3. Hunan Railway Academy Civil Engineering Testing Co., Ltd., Changsha 410075; 4. Hefei Institute of Physical Science, Chinese Academy of Science, Hefei 230031; 5. Guizhou Guijin Expressway Co., Ltd., Guiyang 550081)
Abstract:
In view of the various problems faced by the blasting of mountain tunnels, such as overbreak and under? break and unsuitable muck block size, this paper analyzes the main influencing factors of these problems, and establishes the evaluation index system of tunnel blasting effect. Furthermore, the study establishes a neural network evaluation model for tunnel blasting effects with blasting block size and over- and under-break amount as the prediction target, puts forward an instance segmentation algorithm for the size of tunnel blasting muck blocks based on deep learning, and forms a decision-making method of blasting parameter optimization for mountain tunnels, which is applied to and verified by engineering application. The results show that (1) The size of muck blocks detected by the instance segmentation algorithm of tunnel blasting muck blocks based on deep learning has an error of less than 6.9% (confidence level is 95%) from the true value, realizing the rapid acquisition of the sample data for tunnel muck blocks; (2) The neural network evaluation model of tunnel blasting effect is trained by 148 sets of engineering sample data, after which it can better predict the size of blasting muck blocks and amount of overbreak; (3) The average linear overbreak in the test section after the optimization of the blasting parameters is about 10%, showing an over 50% reduction from the original plan, and the measured size of the muck blocks and amount of overbreak are in good agreement with the model predictions, with a deviation less than 20%.
LEI Mingfeng1 ZHANG Yunbo1 QIN Guifang2 SHI Yuanbo1 GONG Chenjie1,
3
.A Study on Neural Network Evaluation Model of Blasting Effect in Mountain Tunnel and Decision-making Method for Blasting Parameter Optimization[J] MODERN TUNNELLING TECHNOLOGY, 2023,V60(2): 54-61