Inversion Of Surrounding Rock Mechanical Parameters And Displacement Prediction Based On DEPSO-G

The deformation of the surrounding rock in the deep mining roadway is becoming more and more serious,causing huge hidden dangers to the safety of mine production.Taking Jinfeng Gold Mine as the research background,this paper introduces a machine learning algorithm to construct an intelligent displacement inverse analysis model to obtain the mechanical parameters of the surrounding rock and analyze the self-stable ability of the surrounding rock and establish a prediction model for displacement prediction.The main research conclusions are as follows:(1)Create an inverse analysis model as well as a prediction model.Differential evolution particle swarm optimization(DEPSO)is used to optimize the Gaussian algorithm(GP),which solves the problem of the Gaussian algorithm using the conjugate gradient method to select hyperparameters and improves the performance of the GP.(2)In this paper,burial depth,elastic modulus,Poisson’s ratio,cohesion,internal friction angle,and lateral pressure coefficient are selected,and 25 sets of test plans are constructed by the orthogonal design method.As a result,the sensitivity and influence of the parameters on the displacement change were analyzed,and it was concluded that the influence and sensitivity of the burial depth,elastic modulus,and cohesion on the displacement were relatively high.(3)build a displacement back analysis model based on DEPSO-GP using 25 sets of test data as training samples and actual monitoring data from the middle 30 section as test input to obtain the equivalent mechanical parameters of the surrounding rock in the middle section.The relative error between the displacement and the monitored displacement is less than 10%,indicating that the inversion method and parameters are effective.(4)Introducing the strength reduction method,by reducing the internal friction angle and cohesion,when the rock mass reaches its limit state,the reduction coefficient of 1.4obtained at this time is the self-stability coefficient of the roadway,and the roof displacement is obtained.relationship with the reduction factor.(5)The displacement prediction model based on DEPSO-GP has been established,and the average relative error of displacement prediction results is about 2.45%,which indicates that the DEPSO-GP displacement prediction model has high accuracy.(6)In this paper,the data is normalized and converted into logarithms with base e and10.Compared with the prediction results,it is found that the error range of the prediction results after conversion to logarithms with base 10 is the smallest,which proves that the method can effectively improve the Prediction accuracy of the DEPSO-GP model.