search

lidongzh / FwiFlow.jl

Elastic Full Waveform Inversion for Subsurface Flow Problems Using Intrusive Automatic Differentiation
MIT License
40 stars 21 forks source link

Permeability inversion #2

Open acse-ncv19 opened 4 years ago

acse-ncv19 commented 4 years ago

Hi,

Sorry to ask to questions at a time. As I mentioned in my previous question, I am using your code on this website https://lidongzh.github.io/FwiFlow.jl/dev/tutorials/flow/

I also read your paper about Physics Constrained Learning (you reference it on the website above). You explain in the paper that you use a neural network to predict the permeabilities and you also present results (in the paper) that look the same as the ones you the website. When I run your code I do get the same results, but I couldn't find a neural network in your code. Is the neural network built into one of the functions of FwiFlow?

ziyiyin97 commented 3 years ago

I am also a rookie in this field but I think they parameterize their PDE as a neural network, where the coefficients of the PDE (permeability/lame parameters) are regarded as the weights of the network.