Study On The Acceleration Algorithms For Tubulent Combustion Large Eddy Simulations With Detailed Chemical Mechansim

It is an important way to use detailed chemical mechanisms to improve the accuracy of large eddy simulation(LES)of turbulent combustion.However,using detailed mechanisms will lead to time-consuming computation to solve the resulting ordinary differential equation(ODE)of combustion chemistry.On one hand,the large number of species and elementary reactions in the detailed chemical reaction mechanism leads to a large number of chemical ODEs to be solved.On the other hand,the chemical reaction ODEs are stiff and require a stiff ODE solver.Therefore,the present dissertation investigates the characteristics of the acceleration algorithms for the turbulent combustion LES with detailed chemical reaction mechanism from the above two aspects of difficulties.Investigating the acceleration algorithm is beneficial to realize high-fidelity and high-time-efficiency turbulent combustion simulations to guide the design of high-thermal-efficiency low-emission industrial burners.For the difficulty in the large size of detailed chemical reaction mechanisms,mechanism reduction methods can be used to reduce the number of species and elementary reactions.In the present study,the Correlated Dynamic Adaptive Chemistry(CoDAC)is integrated into the in-house combustion numerical simulation code and applied to the numerical simulation of one-dimensional laminar premixed flames and the LES of turbulent jet flame(Sandia Flame-D).The computational accuracy and acceleration effect of the simulation with and without CoDAC are compared,and the characteristics of the CoDAC in turbulent combustion field are analyzed.The results of the one-dimensional laminar premixed flame simulations show that the CoDAC can accurately predict the reaction limit variation characteristics of hydrogen and syngas.The computational accuracy and speed are related to the reduction threshold.The simulation error increases with the increase of the reduction threshold,and the computational time decreases with the increase of the reduction threshold.The results of the LES for the turbulent jet flame show that the CoDAC can accurately predict the profiles of velocity,temperature,important intermediate species and radicals of turbulent jet flame.The accelerating effect of CoDAC is also related to local combustion characteristics.Using CoDAC in LES can reduce the computational time of chemical ODEs by 29%,which is similar to that predicted by the partially stirred reactor(PaSR)with pair-wise mixing(28%),but significantly less than that predicted by the auto-ignition simulation(71%).This is mainly due to the imbalanced computational load of solving chemical reaction in parallel LES and the decreased accelerating effect of CoDAC in the flame reaction zone.Specifically,CoDAC has a better accelerating effect in regions where combustion chemical reaction is less reactive,but poorer accelerating effect in regions where the reaction is more reactive,which aggravates the computational load imbalance among processors and the waste of computational resources.Accelerating stiff ODE solving process can substantially accelerate the solving of chemical reaction problems in combustion simulation.Based on the exponential integrator and Krylov subspace approximation method,an EIKS(Exponential Integrator in Krylov Subspacemethod for combustion simulation is developed and applied to the auto-ignition simulation.The accuracy and acceleration of EIKS are compared with those of the DVODE solver which is widely used combustion simulation.The results show that the accelerating effect of EIKS is substantial compared to the DVODE solver.When coupled with a multi-timescale method and CoDAC reduction,the EIKS can reach a speedup factor of 7.26,which provides a promising application prospect.However,the accuracy of the EIKS is significantly affected by the rounding error.Therefore,further improvements are applied to control the large rounding error of EIKS and an EISKA(Exponential Integrator with Schur-Krylov Approximation)method is developed by introducing Schur decomposition.EISKA is applied to the simulation of PaSR with pair-wise mixing and the LES of turbulent jet flame(Sandia Flame-D).The accuracy and acceleration of EISKA are compared with DVODE solver which is a widely used stiff ODE solver using backward differentiation formula and the accelerating effect of EISKA in different regions of the turbulent combustion field is analyzed.The results show that,compared with DVODE,the EISKA method can accelerate the computation of chemical reactions without the loss of accuracy.For the cases with the improved Li mechanism,GRI-Mech 3.0 mechanism,and USC Mech ?mechanism,the maximum speedup factors of EISKA can reach 1.99,2.61 and 2.19,respectively,with the same level accuracy.In the LES of Sandia Flame-D,the speedup factor of the EISKA method is 2.35 compared with DVODE for the chemical ODE computation and the difference in computational load among processors in parallel computing is reduced.