Research Of Efficient Algorithm Of Isoline Generation In Reservoir Simulation

Visualization is the medium between the results of reservoir simulation and engineers, while isoline is an important representation of visualization. It also plays an important role in meteorology, medicine, earth sciences and many other fields. This paper studies the isoline generation in complex case based on PEBI, isoline generation with multi-core CPU and isoline generation with GPU.In the part of serial isoline generation, this paper firstly discusses the general process of isoline generation, including the calculation of equivalent points, the isoline tracking, smoothing, marking, filling and so on. Then, with the case of faults, this paper presents a method to subdivide the triangular grids near the faults using the PEBI grids. And with the case of area losing near the border caused by directly generating Delaunay triangular grids from PEBI grids, this paper presents a method to generate new triangular grids using the exiting grids. These two methods add only a small amount of grids with a little interpolation, and have little effect with the accuracy of original. The isolines generating match the trends near the faults.In the part of isoline generation with multi-core CPU, this paper firstly analyzes the running time and the parallel possibility of every module in the serial algorithm. Then, it focuses on the isoline tracking which takes the most time. As the isolines are connecting in the adjacent grids, the work area is divided into several parts. The isolines are generated in each part, and then merged to generating the final isolines. The experimental result using OpenMP shows that it can greatly reduce the running time of isoline tracking while running in multi-core CPU.In the part of isoline generation with GPU, this paper firstly analyzes the data structure and the process of isoline tracking. Then, it presents a new merging method as the GPU has huge number of threads. In each round of merging, each thread only takes charge of one pair of equivalent points, and searches the connected points in a part of equivalent points that has already been merged. The search area is increasing in each round, but all the threads are working all the time. The experimental result using CUDA shows that it can greatly reduce the running time of isoline tracking in NVidia GPU.