Engineering lesson
Calibrating heavy end PVT parameters without hiding residuals
A heavy-end calibration should preserve measured targets, parameter bounds and residuals at every state. A fitted interaction parameter is a model adjustment, not a newly measured molecular property.
By PetroCalcHub Team · Updated October 24, 2026 · Free to read and use
Define the fitted target and objective
The supported workflow fits its documented methane/heavy-cut interaction against supplied vapor-fraction observations. The default values are synthetic and explicitly generated, not laboratory measurements.
Retain pressure, temperature and fraction basis for each row. Mixing mass and mole fractions in the target column changes the objective.
Inspect residuals rather than only the fitted coefficient
Observed-versus-predicted pairs and signed residuals reveal trends that an RMSE alone can hide. Bounds and starting assumptions should remain available with the result.
A parameter at a bound can signal limited identifiability or a model mismatch. Numerical termination is not proof that the selected EOS represents the fluid.
Use held-out pressure states
Do not call the fitting observations independent validation. Evaluate additional states or properties that were not used in the objective.
The recorded example reproduces the synthetic fit and preserves its raw targets. It is an implementation demonstration, not field validation of heavy-end characterization.
Explore the example
Actual supported bounded calibration run with its raw target rows preserved in the downloadable record.
Select one of 6 recorded scenarios. Intermediate states are not interpolated.
Pressure (MPa(a)) = 4. Observed vapor mole fraction: 0.729672772; Fitted vapor mole fraction: 0.729663175; Residual: -0.00000959662336.
- Observed vapor mole fraction
- Fitted vapor mole fraction
- Residual
Worked example
Synthetic heavy-end calibration residuals
Default supported gamma heavy-end model: T 350 K, M 200 g/mol, SG 0.85, alpha 2; six synthetic vapor-fraction targets at 1–6 MPa. Full inputs and fitted outputs are downloadable.
Relation and unit basis
residual_i = predicted vapor mole fraction_i − supplied target_i
Targets are synthetic, not laboratory measurements. The fitted interaction is a model parameter, not a measured fluid property.
1. Define the baseline
Pressure (MPa(a)) = 4
Default supported gamma heavy-end model: T 350 K, M 200 g/mol, SG 0.85, alpha 2; six synthetic vapor-fraction targets at 1–6 MPa. Full inputs and fitted outputs are downloadable.
2. Select the relation
residual_i = predicted vapor mole fraction_i − supplied target_i
Targets are synthetic, not laboratory measurements. The fitted interaction is a model parameter, not a measured fluid property.
3. Substitute the values
Fitted methane / heavy-cut interaction parameter = 0.0703360405 dimensionless; Vapor mole-fraction RMSE = 0.0000121872986 dimensionless; Observation count = 6 dimensionless; Vapor-fraction sensitivity norm = 0.202330764 dimensionless
Retain intermediate precision. The displayed values are rounded only after evaluation.
4. Read the result
Observed vapor mole fraction = 0.729672772; Fitted vapor mole fraction = 0.729663175; Residual = -0.00000959662336
Small training residuals verify this synthetic fitting path. Held-out states and independent PVT properties are required for prediction assessment.
Result and interpretation
Observed vapor mole fraction = 0.729672772; Fitted vapor mole fraction = 0.729663175; Residual = -0.00000959662336
Small training residuals verify this synthetic fitting path. Held-out states and independent PVT properties are required for prediction assessment.
The table displays observed and fitted values at recorded pressure states, with residuals retained.
Exact scenario data
Actual supported bounded calibration run with its raw target rows preserved in the downloadable record.
| Pressure (MPa(a)) | Observed vapor mole fraction | Fitted vapor mole fraction | Residual |
|---|---|---|---|
| 1 | 0.8772818 | 0.8772704 | -1.1363e-5 |
| 2 | 0.8228104 | 0.8227937 | -1.6670e-5 |
| 3 | 0.773498 | 0.7734825 | -1.5485e-5 |
| 4 | 0.7296728 | 0.7296632 | -9.5966e-6 |
| 5 | 0.6903039 | 0.6903037 | -1.8867e-7 |
| 6 | 0.6543119 | 0.6543242 | 1.2339e-5 |
Scroll the table to see all columns. The highlighted row is the worked baseline. Downloads retain full numerical precision.
Test your understanding
Work through the question before revealing its answer. Use the stated assumptions and include units in every numerical step.
1. Does near-zero RMSE on synthetic fitting data prove a real-fluid model is accurate?
Show the worked answer
No. It verifies the synthetic calibration path; independent laboratory states and properties are still needed.
2. Reproduce the baseline Observed vapor mole fraction without rounding intermediate values.
Show the worked answer
Fitted methane / heavy-cut interaction parameter = 0.0703360405 dimensionless; Vapor mole-fraction RMSE = 0.0000121872986 dimensionless; Observation count = 6 dimensionless; Vapor-fraction sensitivity norm = 0.202330764 dimensionless. Observed vapor mole fraction = 0.729672772. Small training residuals verify this synthetic fitting path. Held-out states and independent PVT properties are required for prediction assessment.
3. Compare the first and last recorded scenarios. Which input changes?
Show the worked answer
Pressure (MPa(a)) changes from 1 to 6. Other controls remain as stated in the premise. Observed vapor mole fraction changes from 0.8772818 to 0.654311855.
4. What would be needed to apply this example to a different physical system?
Show the worked answer
Replace the synthetic controls with compatible documented inputs and check the relation's assumptions. Targets are synthetic, not laboratory measurements. The fitted interaction is a model parameter, not a measured fluid property. Small training residuals verify this synthetic fitting path. Held-out states and independent PVT properties are required for prediction assessment.
Before using the result
- Retain raw targets and fraction basis.
- Record bounds and objective.
- Use held-out states for prediction checks.
Small training residuals verify this synthetic fitting path. Held-out states and independent PVT properties are required for prediction assessment.
Continue learning
Start with the broader workflow guide
Sources and method
The explanation and synthetic examples are original PetroCalcHub teaching material. The references below support the definitions and model context. They do not supply field validation for the synthetic scenarios.