Research And Application Of Tunnel Blasting Quality Identification Algorithm

With the rapid development of highway,water conservancy,high-speed railway and other engineering construction,more and more tunnels are constructed,and the quality problems such as over excavation and under excavation of tunnel blasting are increasingly prominent,which not only seriously affect the construction safety,but also delay the excavation construction efficiency and greatly increase the project cost.In order to control the overbreak and underbreak well,it is necessary to accurately evaluate the overbreak and underbreak.But at present,the traditional evaluation method of blasting construction quality,especially the overbreak and underbreak of tunnel,seriously lags behind the rapid development of civil engineering technology.Therefore,A method is needed that can accurately identify and quickly determine the quality of tunnel excavation to meet the development needs of current engineering construction.So,relying on the construction of high-speed railway tunnel engineering,the tunnel blasting quality identification algorithm were studied based on deep learning and its application,and achieves a series of results.1.Based on the field investigation and measurement of tunnels such as high-speed railway and highway,it is found that the problem of overbreak and underbreak in tunnel blasting is more prominent,which not only affects the stability of tunnel surrounding rock and structural safety,but also greatly increases the direct and indirect economic costs in the process of tunnel construction.2.A more simple and efficient sensing operation process and data processing method of overbreak and underbreak are proposed,which can collect the point cloud of tunnel contour in the excavation area after field tunnel blasting,and it can quickly and accurately obtain the overbreak and underbreak results and visualize the image.3.A new point cloud data processing method is proposed,which can identify the overbreak and underbreak area,segment the area and label the blasting quality.It can extract high-quality sample data sets.The tunnel blasting overbreak and underbreak method based on deep learning CNN model can be trained well.The overbreak and underbreak images can be segmented,and the overbreak and underbreak quality level can be identified.The results after verification are good,indicating that the model has high reliability.Based on the Yolo algorithm,the original five parameters(x,y,w.H,confidence)were expanded to output 6 basic parameters(x,y,Z,W,h,confidence),which can obtain better training effect.4.Using qt5.9,C + +,python 3.7 and tensorflow2.0 as the programming platform,the perception,recognition and deep learning modules are combined into an organic whole,which can better realize the point cloud visualization,point cloud conversion and overbreak and underbreak point cloud extraction.It can greatly reduce the intensity of overbreak and underbreak measurement,improve the measurement comprehensiveness and accuracy.Thus,it can provide reference to analysis,evaluation and construction control for the overbreak and underbreak quality of long tunnels.So,it may provide an intelligent and efficient technology and method in China.