Research And Application On Intelligent Method Of Surrounding Rock Classification In Highway Tunnels Based On Machine Learning

The classification of tunnel surrounding rock is the basis of tunnel design and construction,which directly affects the safety and operation of tunnels.In order to solve the problems such as difficulty in extracting the actual measured values of parameters and long period for obtaining grading results in the existing surrounding rock classification methods,a Multi-class Softmax Regression(MSR)classification method based on the basic principle of machine learning was constructed in this paper.At the same time,the Python language and Android Studio platform were selected to develop an intelligent classification system of surrounding rock for highway tunnel.And the classification results were compared with the BQ method in three tunnel project examples of Naqiu,Liu Yuetian and Wu Xizuo,which achieved fast and accurate classification of tunnel surrounding rocks.The main research results are as follows:(1)The basic principles of machine learning,classification models and intelligent classification methods for surrounding rocks were systematically analyzed.After comparing the combined multi-class classifier constructed by several binary classification models and other supervised learning algorithm models,the Softmax regression algorithm was selected as the classification prediction model to construct the intelligent classification method;(2)The weight values of each surrounding rock classification index were calculated by using Analytic Hierarchy Process(AHP),and the classification indexes were quantified at the same time.A total of 350 data sets were collected from three tunnels,namely,Naqiu,Liu Yuetian and Wu Xizuo for training and testing the model.The variation for the cross-entropy values of the model in various samples and the classification performance under different learning rates were analyzed;(3)Based on the previous achievements,an intelligent classification system for highway tunnel surrounding rocks was developed,and its functions were extended using Python+Open CV technology based on its debugging and operation to extract rock fissures and obtain quantitative values of rock integrity index;(4)The application of intelligent classification system in actual projects provided a reference for the design and construction of tunnels.Based on comparing the surrounding rock classification specified in the current specifications,the K-fold Cross-validation was used to improve the accuracy of surrounding classification,which is of great significance and value in promoting the development of surrounding rock classification methods.