FieldView13的用户界面
现可支持所有的操作系统平台(windows、Linux、Mac)

-设立了新的标准
FieldView13包含一个新的图形处理引擎。它采用了多核、多CPU设计，利用GPU加速并行计算，多线程和硬件优化以获得所有能使用的核。这个版本使得CFD数据处理更上一层楼，提升的效果与所使用的硬件有关。电脑的配置越好，显像与数据分析的速度及性能的提升越明显。在交互式操作中，本版本比以往任何一个版本都带来至少20倍的速度提升。

默认情况下，FieldView在透明物体的重叠区域会进行高质量的渲染，使其不至于走样。在采用最新技术显卡的计算机上，FieldView能演绎最好的效果。在配置并不怎么好的机器上，它也能正常运行。图形与象素很好地匹配，使得在批量自动运行时也能保持较高的渲染效果。

一个液体燃料射流的直接数值模拟，60亿网格节点描述液体燃料的紊流特性。（Matsuo博士的FieldView图像，JAXA机构）

另一个叫sweep caching的新功能加快了后处理速度，有利于提升用户的使用体验。当用户完成流线、粒子的轨迹线或者瞬态表面的制作，后处理过程中的每一步都记录在内存中。在后续的动画制作过程中，系统快速从内存中提取不断变化的帧。由于动画没有限制，用户可以放大、旋转或者平移动画。

打破障碍
在数据量不断增大的今天，尽管图形处理速度可以提供难以置信的效率改善，但是并不能解决现存的问题。由于高性能计算的普及以及云计算的推广，工程师们不再需要将他们庞大而棘手的模型放在他们的个人电脑上运行，取而代之的是在轻便型笔记本电脑上调用远程桌面，实现全球设计环境。而传统的CFD后处理需要把庞大的计算结果数据通过网络传输过来之后才能打开查看。
“我们当前的计算水平能接受包含上千个时间步长的瞬态模型，每个时间步长都包含20亿网格单元。”Godo经理说：“在年底之前我们能在FieldView13的基础上加以改进，使之能容纳更多的数据。”
FieldView13通过文件压缩使得XDB文件减小，用户可以在原始数据的基础上创建紧凑的子集，这样在调用服务器的数据时可以以最少量的数据来完成分析。XDB文件可以压缩至原始数据的百分之一。这样的话，即使原始数据受现实情况或者FTP的限制，也能在便携电脑上处理XDB文件，如此，一旦数据被打开，它可以被制成动画、切片、探针取值等操作。在计算精度上没有损失，计算速度却大大提高。

V-22Osprey高精度CFD模拟。图为在双CPU/8核工作站上运行FieldView做出的后处理效果图。 多重网格OVERFLOW模拟包含14k网格，668M网格点。（Chaderjian,N.M.和Buning,P.G.的图像，美国直升机协会第67界年度论坛）

整个模型可以生成一个单一的XDB文件，也可以把模型中每一个部件生成相对应的XDB文件。或者，在瞬态分析中，每一个时间步长生成一个同样结构的XDB文件。FieldView13可以同时处理这些文件，并且生成相应的动画与图表。瞬态模型中也可以根据不同的时间步长生成XDB文件，这样可以捕捉到相对较少的缓慢移动片段，同时不影响精度。
XDB文件可以在远端服务器上批量运行，分析者因此可以抽出时间做其它的工作。FieldView把这个过程称之为True Batch Processing。效率不但得到提升，数据安全性也得到提高。数据从未离开过服务器，文件系统也不能手动地进行访问。FieldView13的自动化工具现在已能够使用。

自动提高生产力
FieldView在现有的自动理论基础上实现了XDB文件的批量生成，使CFD数据管理技术上了一个新的台阶。自动化操作的基础就是Restart文件。用户在前一次操作时把设置保存在Restart文件中，那么在下一次创建图片或者表格时可直接调用Restart文件中记录的信息。
FieldView13可通过脚本命令自动处理简单重复的操作。自动操作的最高层次是FieldView FVX，这是一种相对于自然化的程序开发语言，可以完成很多的功能。Intelligent Light公司的开发者认为：用户不需要去学习新出现的或者缩写的计算机语言，所以FVX在使用的时候调用了FieldView的图形用户界面上出现的单词。FVX也可以在现存的Restart文件或者脚本文件中运行，达到更进一步的自动化。
Godo经理证实：在尽量不改变FieldView早先版本的布局与工作流程的前提下，FieldView13有了显著的性能提升，并且100%兼容之前的版本，包括脚本文件。

自行车车轮的非稳态分析
FieldView13的强大之处可以结合一个实际案例来更好地展示。自行车赛车车轮市场的竞争极为激烈，各生产厂家都拿出了高达50万美元的资金用于仿真计算和风洞实验。在稳定性、空气动力学和性能方面有了长足的进步。(来源：NYvelocity.com，“Zipp Lead Engineer Josh Poertner”,2009年8月27日) CFD计算的重要性日益显现，同时CFD能计算的模型结构也日趋复杂。
有一个制造商每周都在空气动力学方面做出检验，瞬态模拟一个高水平骑手在不稳定负荷下所有可能的经历。在一台32核的高性能计算机上，每个时间步长下能生成的原始文件大小为2.8GB，而数值模拟计算通常都需要几百个时间步长。
在一个仅有8核的高性能计算机上可以同时批量运行4个文件的FVX脚本已经研发出来。而从脚本中可以生成Excel文件和XDB文件，其中包含自行车装配中的各个组件，比如：车轮、前叉以及车架。整个XDB文件在每个时间步长大概是60MB，相当于是原始文件的1/46。
此外，利用CFD求解器中后处理器的标准交互式可视界面求解时，相比于单独从每一个时间步长的结果去调用，如果从所有的时间步长调用，我们能花更少的时间去读取XDB文件。
Vince Adams，LMS的客户经理，同时也是仿真计算讲师、咨询顾问以及发言人。他已出版了三部关于FEA的著作并多次在杂志上发表文章。可通过de-editor@deskeng.com联系他。

 

原文参考：

The size of computational fluid dynamics (CFD) models are increasing exponentially as manufacturers recognize the value of detailed transient flow data for temperature, flow resistance and other critical product data. However, one engineer told FieldView Product Manager Matt Godo that he was only able to post-process 1% of his CFD data because of the volume of information and the available time to interpret it. Herein lies the mission driving the developers at Intelligent Light in Rutherford, NJ.

Intelligent Light has been providing products and services for CFD analysts for 27 years. FieldView 13 is the culmination of five years of development--based on customer feedback as well as internal experience from its consulting organization. The release was guided by the following four principles:

1. Revolutionize CFD data management.
2. Provide unprecedented speed and performance.
3. Enable higher productivity sessions and workflows.
4. Don’t require existing users to significantly adapt or adjust the way they have been working.

 

 

The FieldView 13 user interface, now unified for all platforms (Windows, Linux
and Mac).

Setting New Standards
FieldView 13 contains a new graphics processing engine. It uses a many-core, multi-CPU design leveraging GPU acceleration for parallel rendering, multi-threading and hardware optimization to take advantage of all available cores. This release raises the bar for CFD data processing and the benefits scale with the power of the local or client system. The bigger and better the machine, the faster and more powerful the improvements in data visualization and analysis--with interactive operations running at least 20 times faster than the previous release of FieldView.

By default, FieldView displays high-quality renderings of overlapping transparent objects, with anti-aliasing at interactive frames rates. FieldView is designed to deliver the best performance on systems with the latest and most up-to-date graphics cards. It can also be configured to run on systems with more modest graphics resources. High quality rendering extends to batch or automated operation--images generated interactively or in batch are pixel-matched.

 

A direct numerical simulation of a liquid fuel jet with 6 billion gridpoints is
conducted to elucidate the multi-scale, turbulent physics of liquid fuel spray
atomization. (FieldView image courtesy of Dr. Matsuo, Japan Aerospace
Exploration Agency, or JAXA.)

Another new feature that improves post-processing speed, contributing to better interactive experiences for users, is Sweep Caching. When a user completes a streamline or particle path animation or a transient surface sweep, the post-processing objects for each step are stored in memory. Subsequent animations are played from memory for greatly increased frame rates. Free of the limitations of a movie, users can zoom, rotate and/or translate the dataset as it animates.

Breaking Down Barriers
Graphics processing speed can provide incredible improvements in efficiency, but it doesn’t address one of the critical problems facing analysts as datasets increase. With high-performance computing (HPC) clusters and the anticipation of cloud computing, engineers are no longer solving their large, problematic models on their local computer, which is often a laptop in mobile, global design environments. Traditional CFD post-processing requires that results sets in the double-digit gigabyte or even terabyte range be copied across networks before the data can be viewed.

“Our current high-water mark is a transient model covering thousands of time steps, where each step contains over 2 billion cells,” says Godo. “Even larger and more ambitious modeling efforts will be presented before the end of the year, fully relying on features delivered with FieldView 13.”

FieldView 13 addresses this by using reduced representations of the results of interest called Extract Database, or XDB, files. Users can create compact subsets of all the results data of interest on the server system so that a minimal amount of data gets pulled over the network for processing. The XDB file can be as much as 100 times smaller than the raw results dataset. Even if the raw data exceeds physical or realistic transfer or FTP limits, the XDB file will be portable. Then, once the dataset has been opened, it can be animated, sliced, probed and manipulated--with no loss in numerical accuracy, but at greatly enhanced speeds.

 

High-resolution CFD simulation of V-22 Osprey rotor.   Post-processed with
FieldView on dual-CPU/8-core workstation. Multi-grid OVERFLOW simulation
contains 14k grids. 668M grid points. (FieldView 13 image used with permission:
Chaderjian, N. M. and Buning, P.G., “High Resolution Navier-Stokes Simulation
of Rotor Wakes.” Proceedings of the American Helicopter Society 67th Annual
Forum, Virginia Beach, VA. May 3-5, 2011.)

A single XDB file can be created for an entire model, or one can be made for each component or domain in a model. Separate, identically structured XDB files can be created for each step in a parameter sweep, or each time step in a transient analysis. These multiple XDB files can be processed simultaneously in FieldView 13 so that animations or plots across them can be generated. XDB files can also be created using different time steps in a transient model, so that fewer images of slower-moving segments of an event are captured, but resolution on the critical segments can be retained.

XDB files can also be created using batch operations that work remotely on the server, so that the analyst is freed up to work on other tasks. FieldView 13 calls this True Batch Processing. In addition to the productivity enhancements, True Batch also enhances data security. The database never leaves the server on which it was created, and the file system isn’t manually accessed (with all the risk that entails). The automation tools in FieldView 13 come into play here.

Improving Productivity with Automation
The breakthrough in CFD data management through batch creation of XDB files was built on existing automation methodologies in FieldView. The most basic automation option is Restarts. A Restart stores the information needed by FieldView to recreate an image or plot composed by a user when it is saved.

Moving up one level of complexity, FieldView 13 has implemented a scripting capability to automate simple repetitive tasks. At the top of the automation hierarchy is FieldView FVX, a natural language programming tool that provides access to most of the functions within FieldView. The developers at Intelligent Light recognized that users didn’t need to learn new or abbreviated syntax, so the calls within FVX capture words that are used in the FieldView graphical user interface. FVX can also be used to run existing Restarts or Scripts as the basis for more advanced automation.

Godo confirms that the significant gains in performance and productivity were achieved without disrupting the familiar environment and workflows in prior versions of FieldView. Version 13 is 100% backward compatible, he says, including scripts. DE