Research On Risk Assessment Method For Hard Rock Catastrophe In Deep Tunnels

In the condition of deep burial high ground stress,the frequency of spalling damage and rock burst disasters induced by excavation and unloading in hard and brittle rock mass tunnelling has been increasing,which significantly affect the safe construction and long-term stability of tunnels.The risk assessment for hard rock catastrophe in deep tunnels needs to address three key tasks:the evolution mechanism for the existence of risks caused by uncertainties in surrounding rock parameters is not clear,a general risk assessment method for spalling damage has not yet been established,and dynamic risk assessment methods for rock burst under conditions of incomplete information are not yet complete.Based on the method of combining case statistics,theoretical analysis,field monitoring and numerical simulation,risk assessment methods for spalling damage and rock burst based on uncertainty analysis are systematically proposed,which can effectively guide hard rock disaster prevention.The main research results obtained in this paper are as follows:(1)The uncertainty source of three parameters which are uniaxial compressive strength,the ratio of crack initiation and uniaxial compressive strength and tensile strength for the widely used damage initiation and spalling limit approach was analyzed,and their probabilistic statistical characteristics were determined.The multi-output support vector machine(MSVM)was optimized by particle swarm optimization algorithm(PSO),and the intelligent response surface model was established to reflect the nonlinear mapping relationship between back analyzed parameters and multi-source field monitoring data.By combining with the Bayesian analysis method,the B-PSO-MSVM model was established,and the three surrounding rock parameters were dynamically updated simultaneously with the application of the Markov chain Monte Carlo simulation algorithm.The reasons for the difference in the deformation of the tunnel structure before and after the surrounding rock parameters updated were explained,and the evolution mechanism for the existence of risks caused by uncertainties in surrounding rock parameters was revealed.(2)Based on uncertainty analysis of the damage initiation and spalling limit method,a general risk assessment method for spalling damage in hard-rock tunnels is proposed.First,with a reliability-based design method,a probability reliability estimation model for spalling damage depth was established to determine the probability of roof and wall spalling damage based on the stochastic response surface of the Hermite polynomial chaos expansion.Second,the direct economic loss due to damage was calculated by using the spalling damage depth of a unit length tunnel section and the average loss cost;thus,the risk estimation of spalling damage was carried out.Last,a theoretical analytical formula for the spalling damage expected cost ratio was derived to determine probability level thresholds,and the risk grading criteria were calculated.The method was applied to the case of the Canadian deep geologic repository project for low-and intermediate-level nuclear waste.The risk levels of roof spalling damage,wall spalling damage and both types of spalling damage were obtained,and the influences of the commonly used empirical formula for depth estimation,the coefficient of variation and the correlation coefficients of random variables on the estimated probability of spalling damage depth were analysed.(3)Six quantitatively characteristic parameters,namely the maximum tangential stress,the uniaxial compressive strength,the uniaxial tensile strength,the stress coefficient,the rock brittleness coefficient and the elastic energy index were extracted from rock burst cases,and the probability distribution function and correlation of those parameters were determined.The multi-dimensional joint probability distribution function of six characteristic parameters was constructed under the framework of Copula theory,and then the least squares support vector machine(LSSVM)which was optimized by particle swarm optimization algorithm served as the intelligent response surface model to reflect the nonlinear mapping relationship between six parameters and the tunnel rock burst prediction level.Subsequently,the Copula-LSSVM rock burst prediction probability model was established,and the Weibull distribution function of rock burst prediction grade was obtained with the application of the Monte Carlo simulation method.Therefore,a novel tunnel rock burst prediction probability model is proposed.(4)Based on the statistical analysis of rock burst microseismic monitoring cases,six main characteristic parameters,namely,cumulative event number,event rate,logarithm of cumulative released energy,logarithm of energy rate,logarithm of cumulative apparent volume and logarithm of apparent volume rate,were selected from the monitoring cases of rock burst microseismic monitoring to determine the probability distribution function and correlation of these monitoring parameters.A multidimensional joint probability distribution function of the six monitoring parameters was constructed under the copula theoretical framework,and a random forest(RF),optimized by a particle swarm optimization algorithm,was used to create an intelligent response surface model to reflect the nonlinear mapping relationship between the monitoring parameters and the tunnel rock burst prediction level;then,a Copula-RF rock burst prediction probability model was established via Monte Carlo simulation.An equivalent calculation model of the consequence of rock burst accident was established based on the five consequences of casualties,direct economic losses,social impacts,environmental impacts and construction period delays.Rock burst risk classification standards were determined with the resulting risk contour map;thus,a Bayesian network model for rock burst risk assessment was constructed to calculate the probability values of different rock burst risk levels.Therefore,a rock burst risk assessment method based on Bayesian networks was proposed for deeply buried hard rock tunnels.Combined with the possibility estimation of the rock burst risk before construction,a two-stage rock burst risk assessment mechanism was constructed.(5)Systematically analysis was done on the influence of parameter uncertainty and model uncertainty on the prediction results,and the influence of the accuracy and suitability of the Bayesian network model on the evaluation results was analysed.Rock burst cases of the Jinping ? Hydropower Station are used to demonstrate the effectiveness of the proposed method of risk estimation before construction and dynamic risk assessment during construction,and the rationality and practicality of the two-stage risk assessment mechanism are comparative analyzed.