I was wondering if there is an available, real-life known inverse heat transfer problem dataset to benchmark oneselfs algorithm, as in MNIST for deep learning. Talking about... (well in this case I just like bayesian-like approaches)

https://doi.org/10.1080/01457632.2011.525137 https://doi.org/10.1016/j.ijheatmasstransfer.2004.02.028

or thermal diffusivity reconstruction which I like pretty much. I would feel pretty satisfied if I was not creating the problems I'm solving hehe but if I could benchmark against someone. Just want to feel I'm doing something useful.

Thanks for reading!

  • 1
    $\begingroup$ Why not make one yourself and make it publicly available? See related work: open.bu.edu/bitstream/2144/39813/1/MechanicalMNIST_2_20.pdf $\endgroup$
    – NNN
    Commented Jul 10 at 11:49
  • 1
    $\begingroup$ @NNN I could do in what I imagine it is a good inverse problem but I'd really like to know what are the realistic setups: is the noise homoskedastic, white, uncorrelated, how far apart are observations, geometry of the problem, proper priors over thermal conductivity... I can make this stuff up but I'd really like to know what engineers face, would really like to work in realistic numerical scales and numerical error $\endgroup$
    – Aner
    Commented Jul 10 at 12:02


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.