Which Approach Should I Use for Building a Physics-Constrained Model with DeepXDE?

[reactingflow]

Hello everyone!

I am currently learning how to use DeepXDE for training neural network models that include physical constraints. For the one-dimensional system under study, we only have access to observational data at the inlet and outlet under various conditions. Additionally, we are aware of some of the governing constraint equations for this system.

My question is which method would be most suitable for building this model – should I be using Physics-Informed Neural Networks (PINNs), DeepONet, or PI-DeepONet?

Moreover, if I decide to proceed with PINNs, can I train the model using inlet and outlet data collected under different conditions?

Thank you in advance for any guidance you can provide.

Best regards,