How to adjust parameters of both PCSAFT and ideal model with experimental data

[Eduardo-BDMAlves]

I have a dataset of a substance for both density and isobaric heat capacity under high pressure. Looking at the example notebook for fitting data I could not find a way to fit both a model for the PCSAFT equation using Reid model for the ideal part of the equation. In other words is there a way to fit simultaniously PCSAFT coefficients and ReidIdeal model parameters to have a model capable of predicting accurate volume and heat capacities?

[pw0908]

Hi Eduardo,

The way our parameter estimation framework works is by going through the entire model struct and looking through all parameters. As such, if you specify the name of a parameter with its index, even if it’s in a submodel, you can fit it. You could do this for your case.

That being said, I’ve realised a small issue with our implementation of the Reid model. The parameters are stored in a tuple (coefs) for efficiency reasons. I’m not sure if that’ll cause any issues in this particular case. If you end up giving it a try, let us know if you encounter any issues.

I’m away from my laptop at the moment so I’ll give this a shot when I find the time.

[Eduardo-BDMAlves]

Hi Pierre thank you for your answer. Indeed I think I can acces the property given the correct name, I was trying with the name in the documentation (:a,:b,..) however it is transformed to a tuple of coeffs. By passing the right name it seems to access the parameter however it is having problems with it being a Tuple even passing specific indices. It is returning the folowing error:

ERROR: MethodError: Cannot `convert` an object of type Float64 to an object of type NTuple{5, Float64}

I will try with some of the other IdealModels to see if them work. I will post here in case any do!
Is there any easy fix to make the Reid model work with parameter optimization? Would be happy to support if it is not to complicated.

[longemen3000]

I can place back the singular fields a,b,c,d,e in the model, looking back at it, while it does speed up evaluation, it complicates estimation cases

[longemen3000]

in the latest master, the a-e parameters for ReidIdeal were added again. The coeffs field was mantained, but it is recalculated automatically in our estimation framework (or you can call Clapeyron.recombine! if you have a custom-made estimation algorithm)

julia> model = ReidIdeal(["oxygen","nitrogen"])
ReidIdeal with 2 components:
 "oxygen"
 "nitrogen"
Contains parameters: a, b, c, d, e, coeffs, reference_state, Mw

julia> model.params.coeffs
SingleParam{NTuple{5, Float64}}("Reid Coefficients") with 2 components:
 "oxygen" => (30.181499303896256, -0.014916145936966909, 5.47091640274483e-5, -4.9886775708919434e-8, 1.4882888086494297e-11) "nitrogen" => (29.424883205644313, -0.002170074743337995, 5.820123832707267e-7, 1.3053706310500584e-8, -8.231317991971705e-12)

julia> model.params.a[1] = 200
200

julia> Clapeyron.recombine!(model) #function that recalculates coefficients, should work for any model that supports estimation
ReidIdeal with 2 components:
 "oxygen"
 "nitrogen"
Contains parameters: a, b, c, d, e, coeffs, reference_state, Mw

julia> model.params.coeffs
SingleParam{NTuple{5, Float64}}("Reid Coefficients") with 2 components:
 "oxygen" => (200.0, -0.014916145936966909, 5.47091640274483e-5, -4.9886775708919434e-8, 1.4882888086494297e-11)
 "nitrogen" => (29.424883205644313, -0.002170074743337995, 5.820123832707267e-7, 1.3053706310500584e-8, -8.231317991971705e-12)

[Eduardo-BDMAlves]

Thanks for the support, will be using this alteration on master branch.