feos-org / feos-core Goto Github PK

In critical_point_binary, verbosity of VLEOptions is not used - neither in the final computation nor in the Brent optimization. We could add the option into the strut that implements ArgminOp. E.g.:

struct CritOp<U: EosUnit, E: EquationOfState> {
    eos: Rc<E>,
    tp: TPSpec<U>,
    opts: VLEOptions,
}

impl<U: EosUnit, E: EquationOfState> CritOp<U, E> {
    fn new(eos: &Rc<E>, tp: TPSpec<U>, opts: VLEOptions) -> Self {
        Self {
            eos: eos.clone(),
            tp,
            opts
        }
    }
}

yielding

    pub(crate) fn critical_point_binary(
        eos: &Rc<E>,
        tp: TPSpec<U>,
        options: VLEOptions,
    ) -> EosResult<Self>
    where
        QuantityScalar<U>: std::fmt::Display,
    {
        let solver = Brent::new(1e-10, 1.0 - 1e-10, options.tol.unwrap_or(TOL_CRIT_POINT));
        let cost = CritOp::new(eos, tp, options.clone());
        let x = Executor::new(cost, solver, 0.5)
            .max_iters(options.max_iter.unwrap_or(MAX_ITER_CRIT_POINT) as u64)
            .run()?
            .state
            .best_param;
        let moles = arr1(&[x, 1.0 - x]) * U::reference_moles();
        State::critical_point(eos, Some(&moles), None, options)
    }

For a mixture:

cpm = State.critical_point(dft)

returns the error message: RuntimeError: Equation of state is initialized for 2 components while the input specifies 1 components.

The error can be prevented by specifying moles as

cpm = State.critical_point(dft, moles=np.array([1.5, 1.5])*MOL)

A nicer error message could state that specification of moles is required.

Issue to collect typos in docs.

Typos

• "werde used" Parameter.records method

Python

Missing text_signature, typos, etc.

PengRobinsonParameters contains a field joback_records that is never read.

For some equations of state, the initial temperatures are problematic. E.g. for model fluids such as LJ fluids with somewhat sensible parameters critical points can be around 100 Kelvin. For pure substances this is not as much of a problem, since we can provide initial estimates. For binary mixtures, this is not possible. We should either make initial guesses more robust or provide ways to allow initial estimates as input.

In some cases, it would be convenient to directly access (mass) specific densities from a PhaseDiagramPure. Maybe as a getter?

The name of the VLEOptions struct is a relic from when PhaseEquilibrium was still named VLEState. It is appropriate to give it a clearer name, e.g. SolverOptions

and use it in from_json as option type.

The cubic module and PengRobinson equation of state lack docstrings.

Also, to be consistent with other records, PengRobinsonRecord should have a new constructor.

PyO3 0.16.0 provides new features that can be useful for our crate. In particular the added support for simple enums like Contributions and Verbosity can be exploited.

In this example, the state is calculated using N=1 and the given volume instead of calculating the moles from a density iteration.

T = 25*CELSIUS
V = (100*ANGSTROM)**3
p = BAR
z = 3e-4
State(func, T, pressure=p, molefracs=np.array([1-z,z]), volume=V).pressure()
>>> 3.9624kPa

Currently LaTeX is not rendered properly on docs.rs. Missing entry in Cargo.toml?

We should add more information to the README. Although the primary use case for all FeOs crates (from an users standpoint) will be the python package (feos) it might be helpful to have an example on how to use this crate in rust.