基于神经网络的空间流场智能模拟  

ANN-based Intelligent Simulation on Flow Field

作　　者：顾正华[1] 唐洪武[1] 高柱[1] 李云[1] 肖洋[1] 

机构地区：[1]河海大学水利水电工程学院

出　　处：《系统仿真学报》2004年第7期1372-1375,共4页Journal of System Simulation

基　　金：河海大学科技创新基金(2002406443)

摘　　要：基于作者提出的水流智能模型思想,运用人工神经网络方法对透水框架式四面体的尾流场进行智能模拟,为复杂水流的空间流场仿真提供一种新的研究途径。通过水槽实验,借助声学多普勒流速仪(ADV)自动测量分析系统测得四面体九种工况下的空间尾流场,作为四面体水流智能模型的学习样本和测试样本,经过训练获得智能模型的工作表达式,预测流场与实测流场基本吻合。研究结果表明,基于神经网络的水流智能模型,仿真速度快,外延泛化能力和容错能力强,且精度较高,因此具有广阔的应用前景。On the basis of Flow Intelligent Model (FIM) conception from Gu Zheng-hua and Tang Hong-wu, penetrating frame-tetrahedrons wake field was simulated through artificial neural networks method, and a new idea was provided on flow fields simulation. Through flume experiment with Acoustic Doppler Velocimeter (ADV) system, nine kinds of penetrating frame-tetrahedrons wake fields on the different conditions were measured as learning and testing samples of penetrating frame-tetrahedrons FIM. The expression of the wake fields intelligent model was obtained by training, and its simulation result was consonant with prototype. Conclusions illustrate that FIM-based ANN will be prosperous for its rapid simulation speed, good extension generalization capability and tolerance, and high precision.

关 键 词：流场 水流智能模型 智能模拟 透水框架式四面体 人工神经网络 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.9[自动化与计算机技术—控制科学与工程]