10 crucial parameters to check before committing to a CFD software for academia: A comprehensive guide

1. License validity

Anything less than 6 months of license validity is just too short of a duration. This duration corresponds to the length of a semester in most universities. Software installation, geometry preparation, meshing, validation studies, tweaking physics models, post-processing, data analysis and finally arriving at meaningful results is time-consuming, especially if the user is just starting with CFD.

Meshes containing a few million cells are common. You may get away with coarse meshes for small and simple problems. But, that is not applicable when it comes to research or industry. Ideally, there should not be a constraint on the cell count for research, but even if there is, anything less than half a few million cells is just too small. This limit should be even larger for the structured meshes. Talking about structured meshes, they themselves come with some serious limitations when simulating certain categories of problems. That's a topic for another time.

Personal laptops with 16 physical cores are becoming common. University computing clusters can provide you access to hundreds (if not thousands) of cores. Therefore, at the very least the CFD software should be able to run parallelly on at least 16 cores. And it should be scalable.

In case you do not have access to a powerful personal computer or a university cluster, you have cloud computing for your rescue. Therefore, the academic version of the CFD software should be cloud-compatible.

It is difficult to put a number on the size of a healthy user community, but less than 10,000 threads for a multi-physics, general-purpose CFD software is a weak user community in my opinion. 

If at any point there are attempts to involve a sales engineer in the loop who pursues you to buy the CFD software, tries to get details of your research funding, and promises you additional perks for the paid plan, it is best to walk away. They will remain persistent and you will be cornered into buying the software during crucial stages of your research, eventually.

For your research to stay at its best level, it may be important to receive the latest updates with bug fixes, solver changes, accuracy gains, and performance improvements. Ensure that the CFD software provider will provide you with timely software updates.

Most researchers will find that even 6 months of license validity are not sufficient to produce satisfactory results. Therefore, it is important to know if the validity of your license can be extended/renewed in a seamless manner without putting a break in your research if needed. Some of the hassles in renewals could be related to showing extensive proof of your research work in a short span of license validity, or, persistence to buy the license at the end of its expiry.

As per the Journal of Fluids Engineering Editorial Policy on the control of numerical accuracy,

"The Journal of Fluids Engineering will not consider any paper reporting the numerical solution of a fluids engineering problem that fails to address the task of systematic truncation error testing and accuracy estimation. Authors should address the following criteria for assessing numerical uncertainty"

The policy then enumerates 10 criteria/guidelines. Even though the guidelines are for submitting papers to the Journal of Fluids Engineering, they are applicable to any journal submission when using numerical models in your research. The only way to ensure that a CFD software can be trusted is to get hold of its V&V manual.

See if the software provides reasonably accurate solutions for simple and complex flow problems. See if the V&V cases of your interest follow the guidelines stated by the Journal of Fluids Engineering. See if you can re-run the V&V cases on your own computer. You can check OpenFOAM’s V&V online manual for reference. If the CFD software provider does not come with a V&V manual, or, falls short on the recommended numerical analysis practices, look elsewhere.

The extent of customization is definitely the one most severe drawback of commercial CFD software for academic research. CFD software provides varying levels of customization, but none can beat a completely open-source CFD software. Especially if you are a teacher and intend to introduce your students to CFD and numerical modeling, commercial CFD software may not deliver. They operate like back boxes, you can run quick simulations and make pretty figures, but you can’t “learn” CFD using them. 

A CFD software without a powerful and flexible post-processing tool is like a lock without a key. Fortunately, there are many good post-processing tools out there. But, as a researcher, it is important that the format of your CFD results is compatible with the standard file formats such as VTK. The reason is if you start with one CFD software and later on lose access to it (intentionally or unintentionally), you will not be able to open the CFD results for visualization and analysis if they are not in a commonly used format.

Contact Adwaith Gupta at adwaith1990@gmail.com for comments and questions. I constantly guide students in pursuing their career goals in the field of CFD.

Disclaimer: I am not promoting or endorsing any CFD software in this article.