Study On Seed Production Techniquues Of Tibetan Wild Elymus Nutans In Lhasa Valley

Rock burst,also known as rockburst,as a dynamic instability phenomenon of high stress rock mass affected by construction disturbance,is a common dynamic and sudden engineering disaster that is difficult to control and identify in deep geotechnical engineering.Fragile surrounding rock or tunnel vault and other parts of rock mass produce different degrees of cracking,peeling,flaking,sputtering,and strong explosion,which seriously endangers the safety of construction personnel and affects the project progress,Causing huge economic losses.In recent years,with the deep development of plateau mining,transportation,water conservancy and other fields,the ground stress has increased,and the problem of rock burst is increasing.Rock burst has become a key problem to be solved in the construction of deep underground engineering.In addition,the Qinghai-Tibet Plateau is located at the junction of plates,with mountain and canyon landforms widely distributed,the underground stress field structure is extremely complex,and the rock burst phenomenon is very prominent.Therefore,in view of the rock burst problem of the plateau engineering,the establishment of an effective rock burst risk prediction model is the key to the deep geotechnical engineering construction of the plateau.This study combines a variety of optimization algorithms to obtain the optimal parameters of the model and constructs two types of support vector machine based rockburst prediction models.Based on 212 domestic and foreign measured rockburst cases,the model prediction accuracy is used as the identification framework,and the performance of the prediction models is analyzed by integrating the classifier evaluation indexes such as recall rate,and the validity of the models is verified by using both numerical simulation and engineering application.(1)the evolutionary mechanism of rocks is extremely complex,the factors affecting the fragmentation of rocks with non-belief and randomness,for the selection of rock burst evaluation indicators need to take into account a variety of influencing factors.Based on the characteristics and nature of rock burst,this paper selects the maximum tangential stress in the surrounding rock,rock uniaxial compressive strength,rock uniaxial tensile strength,rock stress coefficient,rock brittleness coefficient,elastic energy index as the characteristic indicators of rock burst propensity;taking into account the different classification criteria of the relevant scholars on the level of rock burst,the rock burst is divided into Ⅰ rock burst(minor rock burst),Ⅱ rock burst(moderate rock burst),Ⅲ rock burst(strong rock explosion),Ⅳ rock explosion(severe rock explosion)four levels;through the literature collection method,the 212 groups of rock explosion characteristics of evaluation indicators and rock explosion level one by one correspondence,and finally built a multi-dimensional rock explosion risk evaluation index database.(2)Considering the impact of the actual multidimensional redundant crossover between rockburst evaluation indicators on the model prediction grading performance,the evaluation indicator features are extracted without loss of classification performance by means of principal element analysis.As the actual rock blast grading prediction problem is a complex nonlinear problem,combined with the idea of kernel,the choice of Kernel Principal Component Analysis(KPCA)method for the original rock blast samples to implement the indicator data dimensionality reduction,through the construction of a limited number of independent artificial features,compressed model input indicators,to ensure that most of the information in the original data In the case of the original data to ensure that most of the information,to solve the problem of high-dimensional features,reduce the complexity of the model calculation,the construction of a new system of rock explosion propensity prediction indicators.(3)Using Grey Wolf Optimizer(GWO)and sparrow search algorithm(SSA)to solve the important parameter selection problem of Support Vector Machine(SVM),improve the global optimization capability of the model,and better fit the complex relationship between rockburst evaluation indicators and rockburst level.The GWO-SVM and SSA-SVM rockburst prediction models are established by improving the global optimization capability of the models and better fitting the complex interconnection between rockburst evaluation indicators and rockburst levels.Then,the classifier evaluation indexes such as accuracy,precision,and recall are integrated and compared with the traditional SVM model,PSO-SVM model,and MA-SVM model,and the superiority and reliability of GWO and SSA algorithms over SVM are verified through model validity algorithm experiments.Finally,by integrating 2 optimized support vector machine algorithms(GWO-SVM,SSA-SVM)and KPCA algorithm modeling,two rockburst prediction models,KPCA-GWO-SVM and KPCA-SSA-SVM,are constructed to predict and apply to 8 rockburst examples in Qinling Terminal Tunnel to verify the engineering application value of the 2 prediction models built in this paper.The results show that KPCA-GWO-SVM and KPCA-SSA-SVM can effectively simplify the data structure,have higher prediction accuracy compared with the rest of the models,and the actual engineering prediction results are safer,providing a new method for rockburst propensity prediction.