【作者】 周德闯；

【导师】 汪箭；

【作者基本信息】 中国科学技术大学 ， 安全技术及工程， 2009， 博士

【摘要】 近年来,随着计算机和信息处理技术的快速发展,虚拟现实技术越来越广泛的应用于多个科研领域。本文以构建的PC集群结构的火灾虚拟现实平台为基础,对虚拟现实技术在火灾研究领域中的应用进行了初步的探讨。虚拟现实技术本身涉及到多学科多领域的知识,本文的研究内容涉及的有计算机图形学、数值模拟方法、人工神经网络技术和.NET编程技术几个方面。本文开展的工作如下:首先通过大量的文献调研,论述了在多台PC集群架构下开展虚拟现实仿真所涉及的关键技术。对基于PC集群的多通道可视化系统原理、MPI环境下的场景建模和视景驱动软件的基本理论方法、数据格式、系统结构、模块特性、多通道分布式渲染的实现原理、工作流程及相关的优化技术等内容进行了较全面的论述。系统地论述了粒子系统理论的主要思想和方法,结合火灾场景中火焰和烟气的模拟对粒子系统进行了分析,并以2008北京奥运主火炬燃烧场景仿真为算例,采用数值计算的方法对奥运主火炬的燃烧特征参数和燃烧安全性进行了模拟,分析了燃烧火焰的温度、燃烧对火炬塔的影响、燃烧辐射热对鸟巢顶部钢材和ETFE抗辐射材料的加热效果,计算不仅为奥运主火炬的燃烧安全性评价提供了定量的参考依据,也为进一步的主火炬燃烧的视景仿真提供了粒子系统所需的数据。在此基础上,在Vega Prime仿真开发环境中使用自定义的粒子系统实现了奥运主火炬燃烧火焰和场景的仿真,并完成了在三通道虚拟现实平台上的三维立体呈现。以大空间实验间为研究对象,设计了多种排烟条件的火灾场景,使用火灾动力学的大涡模型对大空间的烟气层降规律和排烟系统效率进行了模拟,分析了排烟风机启动时间、排烟流量、补风口高度、补风口面积和对称性对烟气层降的影响。并以这五个影响因素为输入参数,数值模拟实验数据为训练样本,构造了大空间烟层高度的BP神经网络计算模型,从而为发展快速预测火灾烟气层高度的工程计算模型提供了一种方法。研究了.NET平台与Matlab的混合编程技术及Vega Prime环境下火灾烟气层的可视化和运动控制的方法,本文采用在VC++.NET平台调用matlab引擎的混合编程方式实现了与大空间烟层高度预测模型的数据交互,并以神经网络模型的计算结果数据控制Vega Prime中粒子系统表现的烟气层高度,实现了VR环境下符合真实火灾烟气运动规律的大空间火灾烟气层降的实时仿真。最后,提出了本文研究工作中存在的不足及进一步完善的方法,并对今后将开展的研究内容进行了展望。更多还原

【Abstract】 With the rapid development of computer and information processing technology in recent years,the virtual reality(VR) technology has been more and more widely used in a number of research areas.Based on the fire VR platform constructed by PC cluster,this paper conducted a preliminary discussion on the application of VR technology on fire research.The VR technology refers to multi-disciplinary knowledge in various fields.The study in this paper involved computer graphics, numerical simulation methods,artificial neural network technology and.NET programme etc..The works carried out in this paper are as follows:At first,a big amount of papers about virtual reality technology are investigated.In the current stage of virtual reality development,the systematic framework and systematic working principles applied to virtual reality’s research platform have been illustrated.Furthermore,the concerned key technologies are markedly demonstrated when virtual reality simulation programes are performed at many pc-cluster.And efforts are also made to display the systematic structure,working processes and optimum technology of scenario modeling and simulation driving software in the environment of MPI.The visual simulation of the fire scene in the real environment is virtualized,with the primary focuses on the virtual scenario fire,the particle system and its implementations used in the smoke’s visual simulation.Besides,the Beijing Olympic Games’ cauldron combustion scene poses as an eloquent example.In the designing process of 2008 Beijing Olympics cauldron,the safety and visual impression of the cauldron’s flame is a major concern.First,the combustion scenario of the cauldron is simulated by numerical simulation method.The flame temperature and temperature rise of the cauldron are analyzed.Analysis also goes to the cauldron flame’s influence on steel and ETFE material of the Bird’s Nest building.The computation results provide quantitative data for the safety evaluation of cauldron combustion.Then,the particle system in Vega Prime software is used to simulate the cauldron’s flame and smoke.Finally,the visual simulation of the cauldron’s combustion scenario is accomplished on the three-channel projection VR system.With a big scale space as the research object and with the fire dynamics’ large-eddy model as the basis,the rules of smoke descending is calculated.Meanwhile,BP neural network computing model having the height of big space smoke layer is set up with the help of artificial neural network technology.This research can develop an engineering model which rapidly predicts the height of smoke in the big scale space building like an atrium.Based on artificial neural network technology,with the application of.NET platform and Matlab engine,the smoke movement control methods are studied in the software of Vega Prime.Moreover,the large space’s smoke layer movement in line with the laws of real fire smoke movement is simulated in the environment of Vega Prime.更多还原