Research Of Forecasting Deformation Of The Roadway Rock System

The deformation of roadway surrounding rock can reflect the changeable habitues of the roadway surrounding rock, and provide rational you for judging the stability of rock mass engineering. we can attain the habitues of stability of rock mass advanced by forecasting the deformation, and we can adopt necessary control methods, and insure the stability and safety of rock mass engineering.The aim of this thesis is to resolve the non-linearity in rock mass mechanics that can't be solved by traditional methods and apply artificial intellect and systemic technology to resolve the insufficient parameter and model problem applying numerical methods in rock mass engineering.1) The contents of this study include of these segment:(1) Firstly, feasibility applying neural network methods in the underground engineering is expounded, and secondly model establishment of rock mass underground engineering is expounded, thirdly performances of the neural network are experimented and a rational structure is attained.(2) Rock mass mechanics systems engineering methods are introduced to study regularity of the deformation of roadway surrounding rock and effect factors of deformation of roadway surrounding rock, and analysis relationship of effect factors. These systems methods include of the rock engineering method > the method of partial least-squares' regression, AHP method , the principal component analysis. By these systems method, we can gain the main effect factors for the neural network input factor.2) key problemActual systems mainly is non-linearity systems that can't be established correctly , so the robust of neural network become the main researching problem to improve forecasting precision . the key of this problem is to establish a correct neural network , including selection and sensibility of effect factors(input parameters ), structures of hiding layers, conforming numbers of neural units, transmit function selecting , convergence and stability of neural network and so on .In this thesis the deformation of roadway surrounding rock forecasting problem are analyzed ,and the three main problems for the forecasting are provided. These three problems include of: (1) analysis whether problem can be forecasted; (2) establishing correctly of forecasting neural network model; (3) selection of forecasting problem. In order to solve these three problem, firstly whether the deformation of roadway surrounding rock can be forecasted are analyzed, and insure this problem can be forecasting; secondly problem of data selection is resolved by introducing rock mass mechanics systems engineering methods to study regularity ofthe deformation of roadway surrounding rock and effect factors of deformation of roadway surrounding rock, and analysis relationship of effect factors; at the end rational neural network is established by robust design experiment. In the end a robust neural network is established for forecasting the deformation of roadway surrounding rock, and neural network gains satisfied result.