Hi John, I very much agree with your second point. Data visualization packages are really tools for analyzing results. Without one, you just end up with a bunch of numbers. Of course, you could perhaps integrate the results to get some collective value, but this will not let you see any interesting behavior (flow patters, etc…) that may be occurring in the problem.

Also, the Visualization Toolkit (VTK) is a great tool for performing data analysis. It’s basically a library for processing and visualizing data and it contains all the math and rendering capabilities built in. All you need to do develop your own visualization solution is to couple it with a GUI package such as QT in C or the native Java Swing (and write the code to set up the visualization pipelines). I wrote an article about this if anyone is interested: http://www.particleincell.com/2011/vtk-java-visualization/

Hi Durrell, thank you very much for your contribution on this topic.

I totally agree with your comment on the second trend. More and more simulations, and larger and larger models. Managing and fully utilizing tera or even peta scale simulation data desperately needs a higher-level post-processing tool (a tool that can explore through numerous simulation results to get systematic insights) and simulation data management system.

It is nice to see Tecplot Chorus is tackling this challenge.

Shengwei

Thanks John for your thoughtful input. As you rightly pointed out visualization is just a component (sometimes important; sometimes not) of post-processing, which is always important.

The reality is the increase of CFD licenses sold generally outpaces the increase of full time CFD analyst (which in most company is relatively not changed). So, more and more designer or mechanical engineers are doing a lot of CFD simulations.

They usually expect some sort of turn-key solutions (because CFD simulation is their part time job). This is nothing wrong. The problem is, unfortunately, current CFD software still has very limited intelligence in checking problems in the simulation setups, e.g., are the wall functions valid for the mesh and boundary conditions? Or is the drag law valid for the phases defined and in the multiphase flow regime?

This is the reason why a lot of such new CFD users get frustrated with detailed CFD simulation setup (remember they have other roles). For them, CFD simulation likes a black box.

It is an exciting development that we can see more and more CFD users. The problem is not on these part time CFD users; it is on 1). Slow development in CFD solver intelligence (most of the time, you need a full time analyst to confirm the settings). Most CFD solver vendors like to talk how fast, how accurate, how many models, or now how easy to use for non-CFDers (usually limited to pre- and post-processing tools); and 2). The management often has the misconception that CFD software is so matured (also, not cheap) that it should like a turn-key solution. Engineer just imports CAD model then presses a button. So the management may use these part time CFDers as guerrillas for specific CFD projects.

Shengwei

A second trend we are seeing in the market is the push towards parametric CFD for optimization and performance prediction. This new way of working turns traditional post processing on its head in that engineers need to Plow through 100s of simulations and in the end they model performance using integrated quantities.