Simulation Property Selection for Engineers: Validate Before Tuning
Use steam tables for pure water and steam, an equation of state such as Peng–Robinson (PR) or SRK for hydrocarbons and high reduced temperatures, activity models such as NRTL, UNIQUAC or Wilson for strongly nonideal polar liquids, and a specialized electrolyte model for ionic solutions. Then run a base case, plot Txy or Pxy diagnostics, validate against REFPROP or experimental data, and test the binary interaction parameters that matter most.
TL;DR:
Use a reduced temperature near 0.75 as a guide: above it, favor PR or SRK unless immiscible liquid phases are likely.
Treat UNIFAC and PSRK as initial estimates when parameters are unavailable; for costly design decisions, collect experimental data rather than relying on predictions.
Inspect every binary interaction parameter pair because simulators may silently substitute zero values, then check temperature and pressure limits for each model and parameter set.
Run binary flash tests across the unit’s full temperature range, plot Txy or Pxy diagrams, and compare phase behavior with experiments or REFPROP.
Use sensitivity results to identify which uncertain BIPs affect key outputs, then focus regression or measurement on those parameters instead of expanding the whole model.
Table of Contents
How do you choose a property package with a simple decision tree?
What are the most common property mistakes, and how do you catch them?
How do you validate that your property selection is good enough?
What is the step-by-step workflow from characterization to fixing BIPs?
How do you prioritize which parameters to fix, and where does the data come from?
How does the Jewlz Engineering Toolkit support property work and validation?
How do you choose a property package with a simple decision tree?
Start with five questions, in this order, and stop at the first branch that applies.
Is the system pure water or steam only? Use steam tables or the simulator’s dedicated steam package and skip the rest of the tree.
Is it hydrocarbon or light gas only? Go to an equation of state (EOS), typically PR or SRK.
Does the liquid contain polar or hydrogen-bonding species, or might it show an azeotrope or liquid-liquid split? Go to an activity coefficient model for the liquid phase.
Are dissolved salts, acids or bases dissociating in water? An electrolyte model is required, because generic activity models were not built for ionic speciation.
Is experimental binary interaction parameter (BIP) data available? If yes, use it. If not, a predictive method is your fallback.
When the first four questions leave you undecided, use a reduced temperature check. Estimate it as the system temperature divided by the mole-fraction-weighted sum of the critical temperatures (Kay’s rule). A workhorse model heuristic.pdf) puts the dividing line near a reduced temperature of 0.75: above it, prefer an EOS such as PR or SRK for vapor-liquid equilibrium (VLE) when immiscible liquid phases are unlikely; below it, use a solution (activity) model for the liquid phase and an EOS for the vapor.
Here is a checklist to place your system before you open the simulator:
List every component and flag any that are water, alcohols, acids, amines or salts.
Note the temperature and pressure window the unit will actually see.
Compute an approximate reduced temperature with Kay’s rule.
Check whether the literature reports azeotropes or two liquid phases for your pairs.
Record which BIPs you have data for and which you are guessing.
Property inputs also drive solver stability, so it is worth reviewing our guide to temperature-dependent properties before you commit to a branch.
Which model fits which kind of system?
Think of this as a mapping from chemistry to a sensible starting point, not a final answer—for example, apply control panel cooling calculations when you need to size enclosure ventilation for your equipment.
Water and steam: Use steam tables or your simulator’s dedicated steam package.
Hydrocarbons and light gases: PR or SRK is the usual choice. Simulator documentation maps PR and PRSV to hydrocarbon and high-pressure systems.
Polar or strongly nonideal liquids: Use NRTL, UNIQUAC or Wilson. Reach for predictive UNIFAC only when you have no regressed parameters, and treat its output as a screening estimate.
Electrolytes: Choose a specialized electrolyte model. For ionic systems, consider whether a specialized electrolyte model is appropriate.
Predictive or data-poor systems: UNIFAC or PSRK can give a first pass. If the design decision is expensive, plan a few experimental points instead of trusting a group-contribution estimate.
Dual-model approaches are common for nonideal liquids with a real vapor phase: an activity model handles the liquid, and an EOS handles the vapor. This is the same logic behind the reduced temperature rule above.
How do EOS, activity models and combined methods differ?
An EOS describes pressure-volume-temperature behavior and can be used for gases and hydrocarbons. Assess whether its assumptions fit the phases and operating conditions in your system.
Activity coefficient models take a different route. They describe deviations from ideal mixing in liquids. Check whether a model’s assumptions fit your mixture and whether it is suitable for liquid-liquid equilibrium (LLE) predictions. Consider how to represent the vapor phase separately when using an activity coefficient model.
Combined models such as PSRK and PRSV bridge the gap when both the liquid and the vapor are nonideal. Consider a dual-model setup when your mixture holds polar components but you also run at elevated pressure.
Practical signals help you decide:
Water present, strong hydrogen bonding or known azeotropes: lean toward an activity model.
High pressure, light gases or near-critical conditions: lean toward an EOS.
Both at once: consider a combined or dual-model approach and validate carefully.
The right class is the one whose assumptions match the physics of your dominant phase, not the one with the most parameters.
What are the most common property mistakes, and how do you catch them?
A simulation that converges is not necessarily a simulation that is right. These are the failures we see most often, along with fast tests.
Missing or default BIPs: Many simulators fill gaps with zeros without making noise. Inspect the interaction parameter matrix for every pair and look for blanks or zeros where you expect nonideality.
PR or SRK on strongly polar liquids: The model may predict a single liquid where two exist, or miss an azeotrope. Plot the diagram and compare with literature behavior for the pair.
Ignored applicability ranges: Check the temperature and pressure limits of each model and each parameter set. Watch for unrealistic phase splits or small mass balance oddities at the edges.
Skipping diagnostics: Plot Txy or Pxy and run simple flash cases before trusting any process-level result.
Pro Tip: Before you build the flowsheet, run a binary flash for your two most important components across the full temperature range of the unit. If the phase behavior looks odd there, it will not improve downstream.
How do you validate that your property selection is good enough?
Validation does not have to be exotic. Use a three-step approach: pick a model with system heuristics, run a base case to establish a baseline, then decide whether the model needs improvement by comparing against experimental data or an authoritative database such as NIST REFPROP.
Choose the base model from the decision tree and write down why.
Run the base case and generate diagnostic Txy or Pxy plots and flash results.
Compare to experimental VLE or LLE data, or to REFPROP for the fluids it covers, and decide whether to refine.
Rigor should match project risk. Validation guidance describes a spectrum, from strict empirical validation per ASME VVUQ at one end to weaker expert-review approaches at the other, and recommends planning rigor attributes up front. Physical measurement carries the most weight, while expert review and comparison with another simulation are weaker but sometimes all a low-risk study needs.
Two screening checks are cheap and informative: infinite-dilution activity coefficients for polar pairs, and a handful of flash tests against known behavior. Our write-up on simulation validation and verification covers the broader workflow.
REFPROP is a computational database that provides thermophysical properties for refrigerants, cryogens, hydrocarbons and water, with accuracy often near the limits of available experimental data. For the fluids it covers, it makes a strong reference for judging your simulator’s output.
What is the step-by-step workflow from characterization to fixing BIPs?
Treat this as a checklist you can run in your simulator.
Characterize: Collect Tc, Pc, normal boiling point and acentric factor for the major components, plus expected composition ranges.
Pick: Apply the decision tree and document the chosen model and the assumptions behind it.
Run: Produce Txy or Pxy diagrams and flash separations, and inspect them for azeotropes and liquid-liquid splits.
Test sensitivity: Vary the BIPs you flagged as uncertain. Experiment designs matter here: Latin hypercube or factorial designs explore interactions efficiently when many parameters influence results.
Iterate: Add model complexity only when sensitivity or validation shows it is needed.
That last step follows a “good enough” modeling principle: there is no one-size-fits-all model, validity is tied to purpose and operating context, and the goal is to reduce uncertainty to a level that satisfies project objectives.
Pro Tip: Save the sensitivity results next to the model choice. When a colleague asks why you spent measurement time on one pair and not another, the answer is already on file.
What do the common model names mean in practice?
You will see these names in every simulator menu. Here is a one-line note on each.
Peng–Robinson (PR): A good default for many hydrocarbon and high-pressure applications, and widely supported.
SRK: Similar territory to PR, but check applicability notes when polar species are present.
Lee–Kesler and corresponding-states methods: Useful for crude, reservoir and heavier hydrocarbon systems.
NRTL, UNIQUAC, Wilson: Polar liquids and LLE. When a liquid split is possible, check whether the model you select can represent two liquid phases.
UNIFAC: A subgroup-based predictive option when no regressed parameters exist.
PSRK: Appropriate when vapor nonideality matters alongside nonideal liquids.
Steam tables (IAPWS): The right tool for pure steam and water, ideally through a dedicated steam package.
How do you prioritize which parameters to fix, and where does the data come from?
Do not overhaul the whole database. Use sensitivity results to find the BIPs or pure-component properties that drive your key outputs, and fix those first. Use sensitivity results to identify which parameters drive accuracy, then direct experimental effort there instead of adding model-wide complexity.
REFPROP: It implements Helmholtz-form equations of state, includes transport properties and ships with a DLL for interfacing with other tools. Prefer it over a default EOS when near-experimental accuracy matters and your fluids are covered.
Experimental VLE and LLE: When the sensitivity study says a BIP matters and no good value exists, regress it against measured data.
Documentation: Put property choices, parameter sources and validation diagnostics in your deliverables. Model-building guidance stresses documenting assumptions and performing sensitivity analyses to determine which factors matter.
Why I choose the simplest model that validates
The best thermodynamic model is rarely the most elaborate one. It is the simplest one that still agrees with data where your design decisions live. Complexity without validation buys you more parameters to get wrong and a harder handoff.
Three habits cover most of the damage. Always check for missing BIPs. Plot a Txy or Pxy diagnostic before you trust any flowsheet. Run at least one sensitivity on the BIPs before you freeze the model choice. Then write the ranges and diagnostics into the handoff so the next engineer can retest your assumptions in minutes.
— Joel
How does the Jewlz Engineering Toolkit support property work and validation?
Property selection only pays off if you can check it quickly. The Jewlz Engineering Toolkit offers tools for thermal analysis and simulation.
Visit the engineering toolkit page to learn about licensing and available features. Try the free features, or visit the engineering toolkit page to see how it fits your workflow.
FAQ
What are the main steps in a simulation workflow?
A practical workflow runs from characterization to iteration: collect component properties, pick a base model, run a base case, test sensitivity, and refine only where needed. Validation against experimental data or a reference database closes the loop.
How do I choose properties for a simulation?
Start with the system chemistry: steam tables for pure water, an EOS for hydrocarbons and high reduced temperatures, activity models for polar liquids, and electrolyte models for ionic systems. Then confirm the choice with a base-case diagnostic plot.
Why does missing binary interaction data matter?
Simulators often fill missing BIPs with default values, which can hide real nonideality and produce wrong phase behavior. Inspect the parameter matrix and flag every pair you did not source from data.
Where can I find reliable property data?
For refrigerants, cryogens, hydrocarbons and water, NIST REFPROP is a strong reference, and experimental VLE or LLE literature fills the gaps for other mixtures. Document the source of every parameter you use.
Sources
Recommended
Comments