Examples

Skarstrom ships 33 example decks in examples/decks/, most of them reproducing published adsorption studies. The Python package bundles all of them (skarstrom examples, skarstrom.examples.<name>), so they are the quickest way to see a complete process: run one, inspect how it is set up (skarstrom params deck.json --all), and adapt it.

Every paper-based deck is also an automated validation test; the Validation page lists the simulated and published values and the error metric for each case, including where Skarstrom and the paper disagree, and why.

Example scripts

examples/ has one runnable script per deck family. Each builds its cases from scratch in code (fluid, adsorbents, column, every step and boundary condition, KPIs and solver settings, each value with its unit; build(case) returns the simulation), runs them, prints the KPIs or the cyclic-steady-state status, compares the result with the paper's published values (core/test/papers/*/reference_values.csv) or digitised curves (core/data/) when run from a repository checkout, and saves a plot to examples/output/ when matplotlib is installed (--no-plot to skip). Common options: --case NAME (repeatable) to run a subset, --cycles N for a quick look.

pip install "skarstrom[examples]"        # or pip install . && pip install matplotlib
python examples/ward_2022.py --case 100  # the fastest: about a second

Script	Decks	Process	Compared with	Runtime
ward_2022.py	Ward2022_Q50ccm, _Q100ccm, _Q200ccm	CO₂ breakthrough on activated carbon, three flow rates	digitised outlet y_CO₂ and thermocouple; paper-model breakthrough-time interval	2 s
casas_2012.py	Casas2012_Fig4_*, Casas2012_Fig6_* (3)	CO₂/H₂ breakthrough on AP3-60 at 15 bar	paper-model breakthrough times (Table 5)	8 s
nguyen_2022.py	Nguyen2022_Fig4, _Fig5	CO₂/He breakthrough on CALF-20	digitised outlet y_CO₂ (NRMSE, half-feed crossing time)	1 s
streb_2021.py	Streb2021_A1, _A3	CO₂/CH₄ breakthrough on 13X with thermocouples	digitised outlet mole fractions and five thermocouples	16 s
arora_2021.py	Arora2021_LiLSX_PSA, _PVSA	O₂ PSA and PVSA on LiLSX (50 cycles)	paper-model O₂ purity and net recovery (Table 4)	2 min
ferre_2026.py	Ferre2026_NaYF	Biogas CO₂/CH₄ PVSA on NaY, to CSS	paper-model purity and recovery of both products (Table 5)	4 min
nguyen_2023.py	Nguyen2023_Exp_I … _X (10)	CO₂/N₂ VPSA on CALF-20, with and without LPP, to CSS	measured and paper-model CO₂ purity/recovery (Table 2)	22 min
perez_2019.py	Perez2019_A, _B, _C1, _D1, _E, _F	CO₂/N₂ VSA on 13X, with and without LPP, to CSS	measured and paper-model CO₂ purity/recovery (Table 3)	6 min
vysyaraju_2024.py	Vysyaraju2024_PSA_exp1	He/N₂ 5-step PSA on binderless 5A, to CSS	measured and paper-model He purity/recovery (Table 3)	30 s
ward_pini_2024.py	Ward2024	CO₂/N₂ 4-step VSA on Mg-MOF-74, to CSS	paper-model CO₂ purity/recovery (Table 4)	80 s
tsa_illustrative.py	Illustrative_TSA_13X	Illustrative 4-step TSA with heater and cooler, to CSS	— (KPIs, CSS status, bed temperatures)	3 s
five_step_product_purge.py	FiveStepProductPurge	CO₂/N₂ VSA template with product purge	— (CSS status, integrator statistics)	4 s
optimise_nsga2.py	Perez2019_A	NSGA-II search (pymoo) of feed time and evacuation pressure, each candidate run to CSS in a process pool	— (the CO₂ purity / recovery Pareto front)	about 10 min (estimated: 60 runs of about 30 s, 4 cores)

Runtimes are wall-clock times of a full run of the script (all its cases, to cyclic steady state where the deck checks it) on one core of a shared 4-core Linux machine, for the decks' grid-converged WENO settings and feed-loaded starts (from the paper tests, which run the same cases); for a quick look lower cells or pass --cycles. python/tests/test_examples.py imports every script and runs each once with fast options (one breakthrough case, or two cycles of one cyclic case).

Three executed notebooks are in examples/notebooks/: a parametric study of the feed time of a CO₂ VSA cycle, a sensitivity study with error bars on its KPIs, and isotherm fitting with the engine's fitter.

Worked tutorials

• O₂ PSA (Arora 2021): a four-step Skarstrom cycle for medical oxygen on LiLSX, and how to read purity and recovery when product is recycled.
• CO₂ VPSA with LPP (Perez 2019): a five-step vacuum swing cycle on zeolite 13X with light-product pressurisation, run to cyclic steady state and compared with experiment.
• TSA regeneration: an illustrative temperature swing cycle with heater and cooler energy sources.

Bundled example decks

Family	Decks	What they are
Arora (2021)	Arora2021_LiLSX_PSA, Arora2021_LiLSX_PVSA	O₂ from air on LiLSX zeolite, 4-step PSA and vacuum PSA (Table 4)
Casas (2012)	Casas2012_Fig4_50_50_15bar_25C, Casas2012_Fig6_25_75_15bar_25C, Casas2012_Fig6_75_25_15bar_25C	CO₂/H₂ breakthrough on activated carbon AP3-60 at 15 bar and 25 °C, three feed compositions (Figs. 4 and 6)
Ferré (2026)	Ferre2026_NaYF	CO₂/CH₄ biogas upgrading, 4-step PVSA on zeolite NaY (Table 5)
Nguyen (2022)	Nguyen2022_Fig4, Nguyen2022_Fig5	CO₂/He breakthrough (70 % and 15 % CO₂) into a helium-filled bed of CALF-20 (Figs. 4 and 5)
Nguyen (2023)	Nguyen2023_Exp_I … Nguyen2023_Exp_X	CO₂/N₂ VPSA on structured CALF-20: experiments I–V use the basic 4-step cycle, VI–X the 4-step cycle with light-product pressurisation (Table 2)
Perez (2019)	Perez2019_A, _B, _C1, _D1, _E, _F	CO₂/N₂ VSA on zeolite 13X: A, B and C1 are the basic 4-step cycle, D1, E and F add light-product pressurisation (Table 3)
Streb (2021)	Streb2021_A1, Streb2021_A3	CO₂/CH₄ breakthrough on zeolite 13X with thermocouple data
Vysyaraju (2024)	Vysyaraju2024_PSA_exp1	He/N₂ separation on binderless zeolite 5A, 5-step PSA
Ward (2022)	Ward2022_Q50ccm, Ward2022_Q100ccm, Ward2022_Q200ccm	CO₂ breakthrough on activated carbon at three feed flow rates
Ward (2024)	Ward2024	CO₂/N₂ 4-step VSA on Mg-MOF-74 (Table 4)
Illustrative	Illustrative_TSA_13X	Not a reproduction. A four-step indirectly heated TSA for CO₂ on zeolite 13X, built for the TSA tutorial
Other	FiveStepProductPurge	A CO₂/N₂ VSA cycle with feed, blowdown, evacuation, pressurisation and product purge on a 13X-type isotherm; the feed is split into two sub-steps so that one light-product connection pressurises and the other purges (six steps in the deck). It comes from an integration test, has no KPI configuration and no literature source; use it as a template, not as data

Several decks also carry the experimental curves digitised from their paper (the deck's experimental-data section), the same data as core/data/.

Working with an example

import skarstrom
from skarstrom import ureg as u

path = skarstrom.examples.Perez2019_A.copy_to(".")   # an editable copy
sim = skarstrom.Simulation(path)
print(sim.steps["Feed"])                  # the step's values, with units and string paths
sim.steps["Feed"].time = 80 * u.s
sim.run()
sim.save("Perez2019_A_80s.json")

The cycle templates in data/cycles/ (breakthrough, 4-step VPSA with and without LPP, Skarstrom cycle, 6-step VPSA with rinse) are starting points for building a new process in C++ (see Cycles).