Prediction Of Rockbursts Based On BP Neural Network

Rockburst is a dynamic process under high ground stress conditions.The rapid release of energy causes spontaneous rock explosion,which may result in equipment damage,casualties and construction delays.It is also a very destructive engineering geological disaster,which brings serious losses to the actual project.In the foreseeable future,there is a tendency for China's rock burst prevention and control problems to be increasingly prominent,which will become the significant issue of geological disaster prevention concerning deep underground space engineering.And this results from the continuous advancement of China basic engineering,especially the continuous advancement of the development of underground space and the development of the mining industry.The prediction of rockburst propensity is an important basis for the prevention and control of rockburst disasters.As for the construction of underground space projects and deep ore mining,it is of important theoretical significance and practical value to give feedback on the results immediately,take safety protection measures and avoid losses caused by rock disaster.However,due to the complex mechanism of rockbursts and numerous classification standards,this has made them extremely difficult problems in deep underground construction and mining engineering.In current practice,a variety of rockburst evaluation standards are commonly used,including various internal and external factors that occur in rockbursts,and play important role in rockburst prediction.Rockburst prediction is mainly divided into two categories: long-term prediction and short-term prediction.The main goal of long-term prediction is to provide guidance for decision-making at the initial stage of an engineering project,while the main goal of short-term prediction is to predict the detailed time and location of rock bursts.This article mainly considers long-term rockburst prediction.On the basis of previous studies,it summarizes a large number of practical engineering results and applies them to neural network models,and verifies them with practical engineering experience.The main work and research contents of this article include:(1)Rockburst is an extremely complex dynamic phenomenon.There are many factors that affect rockburst,such as lithology,in-situ stress,rock mass structure,buried depth,groundwater,and construction excavation methods.After comprehensive consideration,this paper selects four of the most important influencing factors,namely uniaxial compressive strength,uniaxial tensile strength,maximum tangential stress and elastic energy index,and performs Spearman correlation analysis to verify that it can indeed be used as Reliable evaluation index.(2)Based on the physical and mechanical characteristics of the surrounding rock reflecting the location of the rockburst and the stress and strain,the rockburst intensity is divided into four levels(ie,no rockburst,slight rockburst,medium rockburst,and strong rockburst).(3)Refer to numerous data and establish a reasonable and reliable sample database.(4)Starting from the principle of multi-objective planning and combining theoretical analysis results,an efficient and accurate back propagation neural network prediction model is constructed.(5)From the perspective of information fusion,a total of 20 engineering examples in three representative regions are used to verify the prediction model,and the best algorithm is selected to train the model.(6)The neural network model is used to predict the rockburst,and various parameter values under various levels of rockburst are obtained.