应用BP神经网络实现基于等高线图像的CFD地形网格
Constructing CFD terrain mesh based on contours image by applying BP neural network
-
摘要: 针对目前生成CFD地形网格时计算量大且精度不高等问题,提出一种基于BP神经网络的CFD地形网格生成方法:将扫描数字化和线状要素提取相结合获取的等高线二维矢量矩阵作为训练样本,生成BP神经网络模型;通过Matlab实现BP神经网络的构建、训练与仿真,拟合出未在等高线上的点的高程值,建立高程DEM栅格矩阵,获得地形三维数据;在网格划分软件Gambit中进行网格生成与优化,生成精细的三维地形网格.实例验证表明,基于BP神经网络模型生成的CFD地形网格精确度高,建模效率也较高,适合CFD模拟工程的应用.Abstract: Aiming at the problem of huge calculation and poor accuracy when generating the CFD terrain mesh,a method for generating CFD terrain mesh based on the BP neural network was put forward,which put 2D vector matrix as the training sample,generating the BP neural network model.The matrix combines with the methods of scanning digitizing and linear elements extraction.And the model was structured, trained, simulated by Matlab.Based on the method, DEM elevation grid matrix was established by fitting the elevation that is not in contours.And then the 3D terrain mesh is generated in the software Gambit based on the data from the matrix.The example validation showed that applying the method can generate high accuracy and efficient CFD terrain mesh, it is suitable for the CFD simulation applications.
-
Key words:
- BP neural network /
- contour image /
- Matlab /
- CFD terrain mesh
-
-
[1]
杨长强,郑永果,郑作亚.利用B样条实现基于等高线的三维地形图[J].信息技术与信息化,2006,37(1):59.
-
[2]
郭晓刚,黄先祥,仲启媛.基于辅助线的等高线生成三维地形算法研究[J].系统仿真学报,2011,23(6):1191.
-
[3]
程雪玲,胡非.复杂地形网格生成研究[J].计算力学学报,2006, 23(3):313.
-
[4]
EI-Fouly T H M, EI-Saadany E F, Salama M M A. Improved grey predictor rolling models for wind power prediction[J]. IET Generation,Transmission and Distribution, 2007, 1(6):928.
-
[5]
张文霄.基于PSO优化的BP神经网络股票预测模型[D].哈尔滨:哈尔滨工业大学,2010.
-
[6]
卢昕昀.基于BP神经网络的超市选址评估研究[D].上海:上海交通大学,2008.
-
[7]
Louka P, Galanis G, Siebert N, et al.Improvements in wind speed forecasts for wind power prediction purposes using Kalman filtering[J].Journal of Wind Engineering and Industrial Aerodynamics,2008,96(12):2348.
-
[8]
蒋兴恒,朱素蓉.基于Levenberg-Marquardt算法改进BP神经网络的卷烟销量预测模型研究[J].经济与管理,2011,17(5):81.
-
[1]
计量
- PDF下载量: 38
- 文章访问数: 1214
- 引证文献数: 0