项目名称: 基于特征的大规模非定常流场可视分析方法研究及软件研制
项目编号: No.11272285
项目类型: 面上项目
立项/批准年度: 2013
项目学科: 数理科学和化学
项目作者: 解利军
作者单位: 浙江大学
项目金额: 78万元
中文摘要: 随着计算能力的提高,越来越多复杂的流体现象可以通过数值方法来模拟,但由此产生的数据往往是大规模、非定常、高维和多属性的,传统的可视化方法难以有效分析这类数据。本项目拟将信息可视化技术和数据挖掘技术应用于基于特征的可视分析方法中,实现大规模非定常流场数据的交互式和智能式分析。具体包括:(1)研究"GPU+多核"架构下的多尺度流场特征抽取、跟踪和度量方法,实现流场特征快速获取。(2)结合信息可视化和科学可视化技术,研究对流场特征进行交互式分析的新方法,增强用户对大规模数据的可视操控能力。具体包括流场几何空间和属性空间的关联分析、高维时变数据在二维和三维空间的映射等。(3)针对大规模非定常流场数据的特殊性,研究高效的数据挖掘算法,自动发掘流场特征的分布、演化和关联关系。(4)将上述算法集成到自主开发的通用可视化软件HEDP/POST中,形成新型的可视化软件,并公开代码供CFD专家使用。
中文关键词: 流场特征;可视分析;信息可视化;科学可视化;流场数据挖掘
英文摘要: With the dramatic increasing computation capacity, more and more complicated fluid flow phenomena can be simulated by using numerical methods. However, large-scale unsteady high-dimensional and multi-variable data generated by means of numerical simulation are hard to be analyzed with the tranditional visualization methods or software. In this proposal, based on flow features, it is planned to combine the information visualization, scientific visualization and data mining technology to analyze large-scale unsteady flow data in an interactive and intelligent way. This study will include four aspects as follows: (1) Multi-scale flow feature extraction, tracking and characterization on the "GPU + Multi Cores" platforms. (2) New visual analyzing method integrating information visualization and scientific visualization technologies. For example, the linking analysis of geometry space and property space, or mapping from high-dimensional time-dependent data to 2D and/or 3D space. (3) Effective data mining algorithms to discover flow patterns by taking advantage of the special properties of unsteady flow data. (4) Implementation of a new visual analyzing software package by integrating the above technologies to HEDP/POST, a general scientific visualization software with independent intellectual property rights. The deve
英文关键词: Flow Feature;Visual Analysis;Information Visualization;Scientific Visualization;Data Mining on Flow Data