PIDeepONet, aux_vars on BC #1704
Replies: 2 comments
-
Beta Was this translation helpful? Give feedback.
0 replies
-
Beta Was this translation helpful? Give feedback.
0 replies
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
-
Dear Pro.Lu:
I encountered some problems while training the model and would like your advice.
Research objectives & Question
Research objectives$[\xi, \eta]\rightarrow[r, z]$ which could transform grids between physical domain coordinates(Cartesian coordinates,$[\xi, \eta]$) and computational domain coordinates(Curvilinear body-fitted coordinates,$[r, z]$). The topography function $z(\xi)$ is the output of DeepNet on the boundary, and the uniform points in compational domain are the input of trunk-net.
Train DeepONet to learning the mapping
Once the model is trained, we want the model to be able to compute physical problems in computational domain according to the chain rule display the results in curvilinear body-fitted coordinates.
Question
what i've tried:
These don't seem to work, I don't understand why DeepONet cannot approximate this boundary condition
bc_topography_top
, do you have any suggestions for this?I tested this issue with another script"stokes_aligned_zcs_pideeponet.py"(after small changes, shown in next block).
Many thanks.
Results&Figures
Loss History
Model predict
Topography function
Code
Beta Was this translation helpful? Give feedback.
All reactions