Parameters Calibration And Conditions Inferences Of Numerical Model Of Mineralization Dynamics Based On Bayesian Optimization

The numerical simulation of ore-forming dynamics is of great significance for the exploration of the genesis of ore deposits,restoration of the ore-forming process,and indication of the favorable areas for deep ore bodies.The mineralization and the formation of ore bodies is the result of the interaction of mechanical deformation,fluid flow,heat transfer and chemical reaction under favorable geological conditions.However,due to the complexity of the ore-forming process and the limitation of the existing geological evidence,there is great uncertainty in the parameters and conditions for solution determination of the dynamic model in the metallogenic process simulation.How to determine the reasonable parameters,initial conditions and boundary conditions is the difficulty of numerical simulation of ore-forming dynamics process.In this regard,based on the Bayesian optimization theory,this paper proposes a method for parameters calibration and conditions inferences of the numerical model of ore-forming dynamics process,aiming to explore the optimal parameters and conditions for solution determination of the numerical model of ore-forming dynamics process by a more scientific way,to restore the ore-forming process and mechanism more objectively,to reveal the formation regularity of ore deposits,and to provide a quantitative basis for metallogenic prediction.This paper takes the Jiaojia gold deposit as the research object,which is in the northwestern Jiaodong gold deposit area,to carry out the parameters calibration and conditions inferences of oreforming numerical simulation.The main results are as follows:(1)On the basis of summarizing the geological background of the Jiaojia gold deposit,the genesis model of the deposit is constructed.The three-dimensional geological modeling of the Jiaojia fault zone was systematically carried out,a three-dimensional structural model of the Jiaojia gold deposit was constructed,and a hexahedral grid model of the three-dimensional structure of the deposit for numerical simulation of oreforming process was constructed.(2)Based on the regional geological data the constitutive equations,momentum conservation and energy conservation relationship of rock mechanics,thermodynamics,fluid mechanics,a coupled numerical model of ore-forming dynamics of the Jiaojia gold deposit was constructed,and then draw up the parameters of the model,discuss how to set the initial and boundary conditions,and determine the parameter search space.(3)Construct a Bayesian model of parameters calibration and conditions inferences of numerical simulation,define the prior probability of the Bayesian model based on the basic knowledge of material parameters and characteristics of metallogenic conditions,and design a Bayesian likelihood function based on actual geological phenomena and numerical simulation results of ore-forming dynamics,to evaluate the degree of fit between the simulation results and actual geological phenomena,and combine the prior probability and likelihood function to obtain a Bayesian optimization objective function that can quantitatively evaluate the rationality of parameter combinations and simulation results.(4)The Bayesian model for parameters calibration and conditions inferences,based on the idea of maximizing the posterior probability,adopts the Tree-structured Parzen Estimation Bayesian optimization algorithm,to realize Bayesian optimization of parameters and conditions of the mechanism-thermal-fluid coupled numerical simulation of the Jiaojia gold deposit,to determine the optimal parameters and conditions of the dynamic model.(5)A comparative analysis was carried out between the simulation results obtained by the optimal parameters and conditions,and the deposit exploration data and other geological evidences.It is found that the simulation results are in good agreement with the actual exploration data,indicating that the Bayesian optimization method is effective.And the optimization results are as follows: the regional stress field is tensile,the surrounding rocks are controlled by an east-west extrusive velocity,and a north-south tensile velocity.Under the dual action of tectonic movement and the property difference between the surrounding rock and the fault zone,violent deformation occurs near the fault zone,and an expansion deformation zone is formed where the primary morphological fluctuation of the fault tectonic transition is obvious,and the rock deformation has changed the rock porosity,permeability,etc.The huge change in properties causes a huge pressure difference near the fault zone.Driven by the pressure difference,the ore-bearing hydrothermal fluid upwells along the fault structure from the deep part.Due to the undulations caused by the deformation and movement of the faults,in the depressions of the fault topography,the fluids converge,the minerals in the ore-bearing fluids aggregate and precipitate to form ore deposits.To sum up,this paper proposes a method for parameters calibration and conditions inferences of ore-forming dynamics numerical model,which is based on Bayesian optimization theory.The efficiency and reliability of this method are proved by a study case,and the method is expected to be useful for 3D ore prospecting prediction and metallogenic target delineation.