DWSIM Thermodynamic Import (prototype)

This notebook shows how to bring compound constants and ideal-gas heat capacities from DWSIM into difflow using difflow.dwsim_import.

DWSIM is a .NET application reached from Python through pythonnet. Its calls are not differentiable, so — like the Cantera and pyglenn importers — this adapter uses DWSIM as a one-time data source and builds difflow’s own JAX-native structures. Unlike pyglenn, DWSIM has critical properties, so it feeds both SpeciesData and CriticalProperties and can build a CubicThermo on its own.

Executable example. DWSIM needs a .NET runtime, so this notebook runs against a small mock implementing the DWSIMBackend contract. All DWSIM contact is isolated in that backend; the import logic and everything downstream is the real difflow code. With DWSIM installed, pass dtl_path=... (or set DWSIM_DTL_PATH) instead of backend=.

import jax
jax.config.update('jax_enable_x64', True)
import jax.numpy as jnp
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline

from difflow.dwsim_import import import_species_data, import_critical_props
from difflow.thermo import IdealThermo, CubicThermo
from difflow.eos import PengRobinson

WARNING:2026-07-14 07:07:43,761:jax._src.xla_bridge:905: Platform 'mps' is experimental and not all JAX functionality may be correctly supported!

1. A DWSIM backend (or a mock)

The real DWSIMBackend wraps DWSIM.Thermodynamics.StandaloneLibrary.dll and exposes constants(name) (MW, Tc, Pc, omega, Tb, Hf in difflow units) and cp_ig_molar(name, T) (J/mol/K). The mock returns the same contract.

# name -> (MW g/mol, Tc K, Pc Pa, omega, Tb K, Hf J/mol, cubic Cp J/mol/K)
_DB = {
    'Methane':        (16.04, 190.6, 4.60e6,  0.011, 111.7, -74600.0,  (33.0, 5.0e-3, 1.5e-5, -6.0e-9)),
    'Carbon dioxide': (44.01, 304.2, 7.38e6,  0.224, 194.7, -393500.0, (22.3, 5.98e-2, -3.5e-5, 7.5e-9)),
    'Water':          (18.02, 647.1, 22.06e6, 0.345, 373.1, -241800.0, (30.5, 9.6e-3, 1.2e-6, -1.1e-9)),
}

class MockDWSIMBackend:
    """Stand-in for DWSIMBackend using difflow-unit values."""
    def list_compounds(self):
        return list(_DB)
    def constants(self, name):
        MW, Tc, Pc, omega, Tb, Hf, _ = _DB[name]
        return dict(MW=MW, Tc=Tc, Pc=Pc, omega=omega, Tb=Tb, Hf=Hf)
    def cp_ig_molar(self, name, T):
        a, b, c, d = _DB[name][6]; T = float(T)
        return a + b*T + c*T**2 + d*T**3
    def close(self):
        pass

try:
    import clr  # pythonnet
    BACKEND = None       # adapter opens a real DWSIMBackend (needs dtl_path/DWSIM_DTL_PATH)
    SOURCE = 'real DWSIM via pythonnet'
except ImportError:
    BACKEND = MockDWSIMBackend()
    SOURCE = 'a mock of the DWSIMBackend contract (install pythonnet + DWSIM for real data)'
print('Data source:', SOURCE)

Data source: a mock of the DWSIMBackend contract (install pythonnet + DWSIM for real data)

2. Import SpeciesData and CriticalProperties

names = ['Methane', 'Carbon dioxide', 'Water']
sp = import_species_data(names, backend=BACKEND)
crit = import_critical_props(names, backend=BACKEND)
for n in names:
    print(f'{n:16s} MW={sp[n].MW:6.2f}  Tc={crit[n].Tc:6.1f} K  '
          f'Pc={crit[n].Pc/1e5:6.2f} bar  omega={crit[n].omega:.3f}  Hf={sp[n].Hf/1000:8.1f} kJ/mol')

Methane          MW= 16.04  Tc= 190.6 K  Pc= 46.00 bar  omega=0.011  Hf=   -74.6 kJ/mol
Carbon dioxide   MW= 44.01  Tc= 304.2 K  Pc= 73.80 bar  omega=0.224  Hf=  -393.5 kJ/mol
Water            MW= 18.02  Tc= 647.1 K  Pc=220.60 bar  omega=0.345  Hf=  -241.8 kJ/mol

3. Build a CubicThermo directly from DWSIM data

Because DWSIM supplies critical properties, one source gives both halves of the real-gas model.

thermo = CubicThermo(IdealThermo(sp), PengRobinson(crit))
T = np.linspace(300, 1000, 60)
fig, ax = plt.subplots(figsize=(6, 4))
for n in names:
    ax.plot(T, [float(thermo.ideal.Cp(n, t)) for t in T], label=n)
ax.set_xlabel('Temperature (K)'); ax.set_ylabel('Ideal-gas $C_p$ (J/mol/K)')
ax.set_title('difflow $C_p$ fit from DWSIM data'); ax.legend()
plt.tight_layout(); plt.show()

flows = {'Methane': 1.0, 'Carbon dioxide': 1.0, 'Water': 1.0}
Tk, P = jnp.array(400.0), jnp.array(5e6)
H_ideal = thermo.stream_enthalpy(flows, Tk, P=None)
H_real = thermo.stream_enthalpy(flows, Tk, phase='vapor', P=P)
print(f'Ideal-gas enthalpy : {float(H_ideal):12.1f} W')
print(f'Real-gas (PR dep.) : {float(H_real):12.1f} W')
print(f'Departure          : {float(H_real - H_ideal):12.1f} W')

Ideal-gas enthalpy :      11150.7 W
Real-gas (PR dep.) :       6175.2 W
Departure          :      -4975.5 W

Summary

• difflow.dwsim_import reads DWSIM compound constants and ideal-gas Cp into difflow SpeciesData + CriticalProperties (one-time; DWSIM is never in the gradient path).

• DWSIM provides critical properties, so it builds a CubicThermo on its own.

• Prototype: all DWSIM contact is isolated in DWSIMBackend; adapt it to your DWSIM version and pass via backend= if needed.