FDS and Ai

[ingdtre]

Hi everyone,
I was googling to find some news about artificial intelligence applied to fire design, I came across this scientific article.

https://www.sciencedirect.com/science/article/pii/S2214157X22007195#sec3

the tool:

http://ifetool.firelabxy.com/atrium

I've been playing around with this online tool http://ifetool.firelabxy.com/atrium that can perform with 97% accuracy for single-story buildings that are regular in plan and height.
I compared them with some of my analyzes carried out over the years, especially for large deposits (which usually have very simple geometries) and the results are superimposable.

For this reason, I kindly ask all of you, is it possible that the FDS team could introduce, on the basis of this new and very powerful technology, the possibility of being able to insert tools for a workflow that is as follows:

1. Objectives of the analysis - identification of the parameters to be evaluated in order to achieve a specific objective;

2. Definition of geometry (in many cases, both professional and study, the geometries we study are simple and regular);

3. Quick and expeditious verification, only for simple geometries (structures such as deposits would be ideal), or carried out with artificial intelligence that can quickly identify the criticalities of the simulation and therefore be able to intervene quickly on the geometry or on the project parameters;

4. CFD analysis

5. Analysis and comparison of results.

In retrospect, CFD analysis would be technically the most valid and above all controllable, but the possibility, in my opinion of being able to have fast, reliable and expeditious results, with AI, would provide professionals with new approaches, still unexplored until now.

Thanks to all.

[Kevin McGrattan]

The "FDS Team" is a handful of people who focus primarily on Step 4, CFD Analysis. The other steps you cite are part of a larger process of fire protection design. I don't know exactly how one would incorporate this "smart" process directly into FDS.

[Jonathan Hodges]

The IFE tool is pretty interesting. It looks like the approach builds on the work we had done in mine fires. We had embedded CFAST predictions of smoke movement through ventilation in the input parameters along with the fire location/size information. It looks like these authors removed those inputs and focused the dataset more on atria. Where I see a tool like this having the most value is either as a plugin to CFAST or Pyrosim where you can rapidly visualize the machine-learning surrogate results while you are building the model. I started to make a plugin for CFAST as part of my pycfasttools toolbox but never quite finished it.

[ingdtre]

Thanks for the reply Kevin, I understand what you are saying and I think that everyone ,every day,  here appreciates the work you do.
Mine was just a starting point to provide any new ideas to the community, but from what Jonathan also replied to me, I think it's something that many are already thinking about.

In answer to Jonathan, yes! I imagine the future looks similar to what you described, probably AI will be incorporated within zone models or in pyrosim to get a quick analysis of what is being modeled.