Bayesian inference of the area profile of a solid in a draining tank using Turing.jl

Specific Domains Probabilistic programming
1

:waving_hand: I am a chemical engineering professor at Oregon State University. we used Turing.jl to solve an inverse problem of reconstruction with quantified uncertainty: infer the horizontal cross-sectional area profile of a solid inside of a tank from the liquid level dynamics as the tank drains. mathematically, this comes down to Bayesian inference of a function involved in a differential equation.

thought to share with you all since
(1) it’s a neat example of applying Bayesian inference to a physical inverse problem
(2) as practitioners, we appreciate how easy Turing.jl was to use for this, and it enabled our research. (thank you, developers.)

14 Likes
2

Thanks for sharing!
Very nice paper indeed. I was looking for something like this.
Did you use Makie.jl for the figures?

Best,
Olivier

1 Like
3

thank you! we used Makie.jl for the figures indeed. (maybe should have cited the Makie.jl paper, too.) we applied a theme from MakieThemes.jl.

1 Like