ADAPTIVE GRID METHOD BASED ON ARTIFICIAL INTELLIGENCE FOR COMPUTATIONAL FLUID DYNAMICS OF NEXT GENERATION

该项目的目的是通过结合人工智能（AT）领域发展的方法，显著地推进自适应网格方法。该项目包括两个阶段。第一阶段，将椭圆方程自适应网格法引入到移动网格系统的有限体积法中。同时，将时空视觉处理的神经网络模型应用于超声速流动数值解中激波和接触面等不连续点的检测。利用神经网络对激波进行自动检测，利用流场信息对激波进行自动检测，将现有的基于神经网络的激波检测方法与自适应网格法相结合。这种检测流场不连续点的方法已被证明是了解流场中流动物理的一种方法。为了检测一般的流动特性，如涡旋、边界层等，已经尝试对该方法进行扩展，但由于神经网络技术的发展，该方法仍处于研究阶段。

The purpose of this project is to progress noticeably a adaptive-grid method by incorporating the approach developed in the field of the Artificial intelligence (AT). The project consists of two phases. At the first phase, the adaptive grid method using elliptic equation is incorporated into a new finite-volume method in moving grid system. At the same time, a neural network model of spatiotemporal visual processing has been applied to detect discontinuities, such as shock wave and contact surfaces, in a numerical solution of a supersonic flow. The present shock wave detection method using neural network approach has been successfully incorporated into the adaptive grid method since the shock waves are automatically detected by neural network and using the information of flow field.The present approach to detect discontinuities in the flow field has been proved surely a way to know the flow physics in the flow field. the method has been tried to extend in order to detect general flow properties, such as vortex, boundary layer and so on, and, however, it is still under studies since the neural network technology is also developing.