Near-wall Blood Flow Modeling with Physics-Informed Neural Network (PINN).

[@billyboy2SNU]

This is a great presentation about PINN model application.
recently, I have been deeply interested in this model for the water flow simulation.
then I felt big barriers for the input data construction for the model.
Can I get tips for the input geometry data set like mesh generation for the CFD model? or any recommendation ?

[@nikhilyewale2639]

Has the effect of sensor locations been studied ?

[@sarasooren2673]

Dear Professor thank very much for share Code and video, Would mind tell what is the size of channel? I think channel size is whether nano or micro ?

[@mramirh1]

It is all non-dimensional. The Reynolds number is reported in the paper.

[@sarasooren2673]

@@mramirh1 I suppose behavior of fluid in nano or micro is diffrent form CM size! I am right !!!!

[@bharathmadduri]

Arzhani, this is my second question.
I am trying to implement your method to predict velocity and pressure fields in a spinning shaft which is partially filled with oil, and outlets exposed to air, from 10 experimental data points.
In traditional CFD I use the multiphase method, with close BC/IC as experiment.
Can I use PINN on a spinning shaft, with oil sloshing everywhere and multiphase flow. Can PINN be used to measure both air and oil velocity / pressure fields
Please let me know. Thank you

[@mramirh1]

If you have a well-posed problem (meaning you know all BCs and parameters) then traditional CFD method is much more robust and faster. PINN is good when you have some measurement data and some unknown parameters in the CFD model.

[@bharathmadduri]

@@mramirh1 thank you 🙏

[@bharathmadduri]

Hi Arzhani. Great presentation. My question
1) in which software did you code PINN
2) How did the python-pytorch code understand wall location

[@mramirh1]

Pytorch. By using VTK libraries boundary mesh data in VTK format can be extracted. You can see the code here: github.com/amir-cardiolab/PINN-wss

[@bharathmadduri]

@@mramirh1 thank you 👏