I'm working on a fluid simulation prediction task using an FNO3D model and would like to verify the feasibility of my approach. Here are the key details:

​​Task Objective​​:
Predict 30 future frames of density fields (256×256) given:

First 10 frames of density maps (density_t0-t9)
Physical constants:
Geometry dimensions (parametric)
Motion frequency (Hz)
Motion amplitude (m)
Non-fluid regions are masked as 0 in all frames

Is the FNO3D architecture suitable for this spatiotemporal prediction task with parametric physical constraints?