A Study Of Parallel Chemistry Acceleration Algorithms And Applications In The Gaseous Detonation Simulations

The self-sustaining propagation of gaseous detonation wave is essentially one of transient and compressive reacting flows,which involves in a series of complicated physical and chemical mechanisms such as reaction kinetics,reaction thermodynamics and hydrodynamics.Since it is an important phenomenon in the fields of hypersonic propulsion and industrial explosion events,the relevant study has scientific and pratical significance.From the view of numerical simulation,there are a large number of chemical species and elementary reactions in the gaseous detonation computations with detailed chemical reaction mechanism.Due to the extreme difference of time scales among these species,serious stiffness and nonlinearity exist and further cause the expensively computational cost when the chemical source term is directly integrated.In order to improve the computational efficiency of chemical reactions,a series of parallel chemistry acceleration algorithms based on the reduced in situ adaptive tabulation(ISAT)method are proposed for simulations of transient compressible reacting flows.These algorithms are then applied in the simulations of gaseous detonation propagation to identify the performances on both computational accuracy and efficiency.The main work and the corresponding results are as follows:Firstly,a series of parallel chemistry acceleration algorithms are developed based on the table operations.The algorithms are formed by combinations of two types of parallel strategies,one is called as primary strategy which includes Purely Local Processing(PLP)and Transposed Processing(TP),another is called as balanced strategy which contains Data Apportion Processing(DAP)and Data Exchange Processing(DEP).Both types of parallel strategies are proposed based on the distribution of thermodynamic states in detonation flow field.The primary strategy can form an algorithm either separately or in combination with balanced strategy.Thus,in according to the primary strategy,these algorithms are divided into two groups,that is,PLP-based algorithms and TP-based algorithms.Compared to the computational results by direct integration(DI),all the parallel chemistry acceleration algorithms can improve the computational efficiency of chemical reactions without the loss of computational accuracy.The maximum speedup ratio of 4.60 is obtained for the optimal algorithm.To explain the difference of acceleration effects among the different algorithms,two indexes namely load balance and synchronization of table operations(including query,enlargement,insertion,retrieval and deletion of nodes in a table)are proposed.It is found that the load balance and synchronization of table operations among different data tables both have the significant influence on the speedup ratio of our algorithms.Concretely,good load balance means that the computational loads among different tables are reasonable and equivalent;while good synchronization means that the specific operations among different tables are performed at the same pace.In the simulations of two-dimensional gaseous detonation propagation,the algorithm with higher speedup ratio usually has better performances on both load balance and synchronization among table operations.Secondly,the effects of chemical reaction mechanism(fuel type)and numerical scheme resolution on the performance of parallel chemistry acceleration algorithm are analyzed.The TP/DEP algorithm is selected and applied in the simulations of two-dimensional gaseous detonation wave with different conditions.Results show that the TP/DEP algorithm can provide the satisfactory computational accuracy,while the computational efficiency of it depends on different conditions.The chemical reaction mechanism has a more obvious influence on the computational efficiency,while the numerical scheme resolution has a relatively unimportant effect during the entire simulations.Furthermore,the load balance and synchronization among table operations for different cases are analyzed.It is found that both of them jointly cause the difference of speedup ratios.Among them,the load balance of table operations can be improved through the appropriate selection on parallel strategies(e.g.TP and DEP),while the improvement of synchronization is more complex.For the purpose of improving the synchronization among table operations,a size control strategy is proposed to maintain the size of data tables in the computations.This table size control strategy is then combined with the selected TP/DEP algorithm.Two different methods namely single table size control and total table size control are adopted in this strategy.The single table size in the former represents the upper limit of node number of each single table,while the total table size in the latter is the upper limit of total node number of all tables in computations.In the numerical simulations of two-dimensional gaseous detonation wave propagation,the TP/DEP algorithm enhanced with the table size control strategy is examined to be accurate in all computational cases with different sets on both single table size and total table size,while the speedup performance among cases are different.Furthermore,our correlation analysis indicates that an optimization range of[1.3,2.2]is exists in the proportionality coefficient constructed by the single table size,the total table size and the number of sub-zones.Within this range,the use of table size control strategy can improve the load balance and synchronization among table operations at the same time and thus give rise to the speedup ratio.In this case,a principle for determining the single table size and the total table size in chemistry acceleration computations is obtained,which is helpful for uses to choose the appropriate table size.Since the computer memory and the number of sub-zones are both known before the computations,the total table size must be firstly determined to make full use of the memory space,while the single table size is then be determined through the relation within the optimization range of[1.3,2.2].Lastly,the parallel chemistry acceleration algorithms enhanced with the table size control strategy is applied into the numerical simulations of two-dimensional 2H₂+O₂ oblique detonation,the effects of both parallel strategy selection and parallel sub-zone number are analyzed.All the algorithms show the excellent performances on computational accuracy when compared with the results of DI.As for the computational efficiency,the selection on parallel strategy and the number of parallel sub-zones are found to have the obvious influence on speedup ratio.In the matter of strategy selection,the acceleration effect of single PLP algorithm is worse than those of PLP/DAP and PLP/DEP in the PLP-based algorithms;while the acceleration effect of single TP algorithm is better than those of TP/DAP and TP/DEP in the TP-based algorithms.As for sub-zone number,the reduction on sub-zone number can improve the load balance and synchronization among table operations and thus give rise to the increase of speedup ratio.At the end of this paper,the general principle for choosing and setting the parallel chemistry acceleration algorithm with the table size control strategy is proposed,which is helpful for users in the numerical simulations of transient compressible reacting flows.