Application Of Parallel Computing In Environmental Engineering

Since ancient times,groundwater resources are the most precious wealth of mankind.It is the main source of domestic water and drinking water for human life.In recent years,due to the continuous progress of society and the rapid development of industrialization,human beings have begun to over-exploitation of water resources,and coupled with unreasonable planning and utilization,frequent water pollution,making China's groundwater resources become increasingly scarce.Therefore,the rational use andmanagement of groundwater resources are imperative.The numerical simulation of groundwater flow is an effective method to realize the evaluation and monitoring of water resources in a certain area,so that the groundwater resources can be supplemented and restored periodically.With the deepening of the research on groundwater,the increasing complexity and scale of the research,the ordinary serial algorithm can not meet the needs of the researchers in calculation time,and the huge amount of data is the huge consumption of computer memory.The Parallel processing technology is a good solution to this problem.The main work is the parallel processing of numerical simulation of groundwater flow.In order to solve the large-scale partial differential equations appearing in the groundwater simulation,we can solve the existing problems of groundwater simulation by combining the theoretical knowledge of the advanced solution equations in the numerical field,and the parallel calculation method.Shorten the time and memory consumption of numerical simulation.This topic is mainly based on the original model of modflow,so as to solve the large-scale linear equations as the main research content,mainly to shorten the running time as the main purpose,focusing on the mainstream of the parallel structure system and algorithm ideas.The main work includes: 1)Study the realistic meaning and theoretical basis of the three-dimensional finite difference formula mathematical model to understand its overall model architecture and the concept and program expression of each basic subroutine,and compile and run it programmatically.2)In this paper,we study the iterative algorithm used in large linear equations to study the idea and principle of preprocessing conjugate gradient method and parameter estimation method,and to ensure that the two algorithms are parallelized with different parallel techniques.3)For the parallel architecture,the parallel distributed memory MPI parallel and shared memory OpenMP parallel and GPU-based parallel CUDA parallel principle,with OpenMP and OpenMP / MPI two parallel methods for simulation test,the experimental results show that the two parallel algorithms can significantly improve the running rate,and within a certain range,the efficiency becomes more and more obvious as the scale increases.