Asphaltene Parameter Fitting

Overview

The AsphalteneOnsetFitting class provides automated CPA parameter tuning using the Levenberg-Marquardt optimization algorithm. This enables matching CPA model predictions to experimental asphaltene onset pressure (AOP) measurements.

Why Parameter Fitting?

CPA parameters for asphaltenes (association energy ε/R and volume κ) vary between crude oils and must be tuned to experimental data for accurate predictions:

Parameter	Effect on Onset Pressure
Higher ε/R	Earlier precipitation (higher onset P)
Higher κ	Stronger aggregation tendency
Higher MW	Reduced solubility

Default parameters provide reasonable estimates, but tuning to measured AOP data improves prediction accuracy significantly.

Quick Start

import neqsim.pvtsimulation.util.parameterfitting.AsphalteneOnsetFitting;
import neqsim.thermo.system.SystemSrkCPAstatoil;

// 1. Create fluid with asphaltene
SystemSrkCPAstatoil fluid = new SystemSrkCPAstatoil(373.15, 200.0);
fluid.addComponent("methane", 0.30);
fluid.addTBPfraction("C7", 0.50, 0.150, 0.80);
fluid.addComponent("asphaltene", 0.05);
fluid.setMixingRule("classic");

// 2. Create fitter and add experimental data
AsphalteneOnsetFitting fitter = new AsphalteneOnsetFitting(fluid);
fitter.addOnsetPointCelsius(80.0, 350.0);   // T=80°C, P=350 bar
fitter.addOnsetPointCelsius(100.0, 320.0);  // T=100°C, P=320 bar
fitter.addOnsetPointCelsius(120.0, 280.0);  // T=120°C, P=280 bar

// 3. Set initial guesses and run fitting
fitter.setInitialGuess(3500.0, 0.005);  // ε/R [K], κ [-]
boolean success = fitter.solve();

// 4. Get results
if (success) {
    System.out.println("Fitted ε/R: " + fitter.getFittedAssociationEnergy() + " K");
    System.out.println("Fitted κ: " + fitter.getFittedAssociationVolume());
}

Detailed API

Constructor

// Create fitter from existing fluid system
AsphalteneOnsetFitting fitter = new AsphalteneOnsetFitting(fluid);

The fluid system must:

Be a CPA system (e.g., SystemSrkCPAstatoil)
Contain an asphaltene component
Have a mixing rule set

Adding Experimental Data

// Add point with temperature in Kelvin
fitter.addOnsetPoint(373.15, 320.0);  // T [K], P [bar]

// Add point with temperature in Celsius (convenience method)
fitter.addOnsetPointCelsius(100.0, 320.0);  // T [°C], P [bar]

// Add point with measurement uncertainty
fitter.addOnsetPoint(373.15, 320.0, 10.0);  // T [K], P [bar], stdDev [bar]

// Clear all data points
fitter.clearData();

Recommended: Use at least 3 data points at different temperatures for robust fitting.

Setting Initial Guesses

// Two parameters: association energy and volume
fitter.setInitialGuess(3500.0, 0.005);  // ε/R [K], κ [-]

// Single parameter (when fitting only one)
fitter.setInitialGuess(3500.0);

Recommended Initial Guesses

Oil Type	API Gravity	ε/R [K]	κ
Light oil	>35°	2500-3500	0.005-0.015
Medium oil	25-35°	3000-4000	0.003-0.008
Heavy oil	<25°	3500-4500	0.002-0.005

Fitting Parameter Types

Control which parameters are fitted:

import neqsim.pvtsimulation.util.parameterfitting.AsphalteneOnsetFunction.FittingParameterType;

// Fit both association parameters (default)
fitter.setParameterType(FittingParameterType.ASSOCIATION_PARAMETERS);
fitter.setInitialGuess(3500.0, 0.005);

// Fit only association energy
fitter.setParameterType(FittingParameterType.ASSOCIATION_ENERGY);
fitter.setInitialGuess(3500.0);

// Fit only association volume
fitter.setParameterType(FittingParameterType.ASSOCIATION_VOLUME);
fitter.setInitialGuess(0.005);

// Fit binary interaction parameter
fitter.setParameterType(FittingParameterType.BINARY_INTERACTION);
fitter.setInitialGuess(0.0);

// Fit molar mass
fitter.setParameterType(FittingParameterType.MOLAR_MASS);
fitter.setInitialGuess(750.0);

Configuring Pressure Search

// Set pressure range for onset calculation
fitter.setPressureRange(
    500.0,  // Start pressure [bar] - search starts here
    10.0,   // Min pressure [bar] - search stops here
    10.0    // Pressure step [bar] - coarse search step
);

Running the Fit

boolean success = fitter.solve();

if (success) {
    // Fitting converged
    double[] params = fitter.getFittedParameters();
} else {
    // Fitting failed - check initial guesses or data quality
}

Getting Results

// Get fitted parameters
double epsilonR = fitter.getFittedAssociationEnergy();
double kappa = fitter.getFittedAssociationVolume();

// Get all fitted parameters as array
double[] params = fitter.getFittedParameters();

// Check if fitting was successful
boolean success = fitter.isSolved();

Predicting Onset at New Conditions

// Calculate onset pressure at a new temperature
double onsetP = fitter.calculateOnsetPressure(393.15);  // T [K]

// Get the tuned fluid system
SystemInterface tunedFluid = fitter.getTunedSystem();

Algorithm Details

Levenberg-Marquardt Optimization

The fitting uses the Levenberg-Marquardt algorithm to minimize:

\[\chi^2 = \sum_{i=1}^{n} \frac{(P_{\text{calc},i} - P_{\text{exp},i})^2}{\sigma_i^2}\]

Where:

$P_{\text{calc},i}$ = calculated onset pressure at temperature $T_i$
$P_{\text{exp},i}$ = measured onset pressure at temperature $T_i$
$\sigma_i$ = measurement uncertainty

Convergence Criteria

Maximum iterations: 50 (configurable)
Parameter tolerance: 1e-6
Function tolerance: 1e-8

Onset Pressure Calculation

For each temperature, the onset pressure is found by:

Starting at high pressure (single liquid phase)
Decreasing pressure in steps
Detecting when a second phase appears
Bisecting to find exact onset point

Best Practices

Data Quality

✅ Use at least 3 onset points at different temperatures
✅ Include temperature range spanning expected field conditions
✅ Use reliable measurement techniques (HPM, light scattering)
✅ Specify measurement uncertainty for weighted fitting

Initial Guesses

✅ Start with literature values for similar oils
✅ Use the “heavy” guess for conservative prediction
✅ Try multiple initial guesses if fitting fails

Validation

✅ Check that fitted parameters are physically reasonable
✅ Validate predictions at conditions not used in fitting
✅ Compare with De Boer screening for consistency

Troubleshooting

Fitting Fails to Converge

Possible causes:

Poor initial guesses
Inconsistent experimental data
Fluid system not properly initialized

Solutions:

Try different initial guesses
Check experimental data consistency
Ensure mixing rule is set: fluid.setMixingRule("classic")

Fitted Parameters Outside Physical Range

Expected ranges:

ε/R: 2000-5000 K
κ: 0.001-0.1

If outside range:

Check experimental data quality
Consider simpler fitting (single parameter)
Review fluid composition

Poor Prediction Accuracy

Possible causes:

Limited temperature range in training data
Extrapolating beyond calibration range
Composition effects not captured

Solutions:

Add more experimental data points
Limit predictions to calibrated range
Consider refitting for different conditions

Example: Complete Fitting Workflow

import neqsim.pvtsimulation.util.parameterfitting.AsphalteneOnsetFitting;
import neqsim.thermo.system.SystemSrkCPAstatoil;

public class AsphalteneParameterTuning {
    public static void main(String[] args) {
        // Create realistic oil composition
        SystemSrkCPAstatoil fluid = new SystemSrkCPAstatoil(373.15, 350.0);
        fluid.addComponent("methane", 0.35);
        fluid.addComponent("ethane", 0.08);
        fluid.addComponent("propane", 0.05);
        fluid.addTBPfraction("C7", 0.35, 0.100, 0.75);
        fluid.addTBPfraction("C15", 0.12, 0.200, 0.85);
        fluid.addComponent("asphaltene", 0.05);
        fluid.setMixingRule("classic");
        fluid.init(0);
        
        // Create fitter
        AsphalteneOnsetFitting fitter = new AsphalteneOnsetFitting(fluid);
        
        // Add experimental AOP data from lab tests
        // (Temperature in Celsius, Pressure in bar)
        fitter.addOnsetPointCelsius(60.0, 380.0);   // Cool conditions
        fitter.addOnsetPointCelsius(80.0, 350.0);   // Intermediate
        fitter.addOnsetPointCelsius(100.0, 320.0);  // Reservoir T
        fitter.addOnsetPointCelsius(120.0, 280.0);  // Hot conditions
        
        // Set initial guesses (medium oil)
        fitter.setInitialGuess(3500.0, 0.005);
        
        // Configure pressure search range
        fitter.setPressureRange(500.0, 10.0, 10.0);
        
        // Run fitting
        System.out.println("Starting parameter fitting...");
        boolean success = fitter.solve();
        
        if (success) {
            System.out.println("\n=== FITTING RESULTS ===");
            System.out.printf("Fitted ε/R: %.1f K%n", fitter.getFittedAssociationEnergy());
            System.out.printf("Fitted κ: %.5f%n", fitter.getFittedAssociationVolume());
            
            // Validate: predict onset at test temperatures
            System.out.println("\n=== VALIDATION ===");
            double[] testTemps = {333.15, 353.15, 373.15, 393.15};  // 60-120°C
            double[] measuredP = {380.0, 350.0, 320.0, 280.0};
            
            for (int i = 0; i < testTemps.length; i++) {
                double predP = fitter.calculateOnsetPressure(testTemps[i]);
                double error = predP - measuredP[i];
                System.out.printf("T = %.0f°C: Measured = %.0f bar, Predicted = %.1f bar (Δ = %.1f)%n",
                    testTemps[i] - 273.15, measuredP[i], predP, error);
            }
            
            // Predict at new conditions
            System.out.println("\n=== PREDICTIONS ===");
            double newOnset = fitter.calculateOnsetPressure(413.15);  // 140°C
            System.out.printf("Predicted AOP at 140°C: %.1f bar%n", newOnset);
        } else {
            System.out.println("Fitting failed - try different initial guesses");
        }
    }
}

References

Li, Z., and Firoozabadi, A. (2010). “Modeling Asphaltene Precipitation by n-Alkanes from Heavy Oils and Bitumens Using Cubic-Plus-Association Equation of State.” Energy & Fuels, 24, 1106-1113.
Vargas, F.M., Gonzalez, D.L., Hirasaki, G.J., and Chapman, W.G. (2009). “Modeling Asphaltene Phase Behavior in Crude Oil Systems Using the Perturbed Chain Form of the Statistical Associating Fluid Theory (PC-SAFT) Equation of State.” Energy & Fuels, 23, 1140-1146.
Gonzalez, D.L., Ting, P.D., Hirasaki, G.J., and Chapman, W.G. (2005). “Prediction of Asphaltene Instability under Gas Injection with the PC-SAFT Equation of State.” Energy & Fuels, 19, 1230-1234.

Asphaltene Parameter Fitting

Comprehensive Java library for thermodynamic, physical property, and process simulation. NeqSim provides tools for PVT analysis, phase equilibria, transport properties, and full process simulation.

Asphaltene Parameter Fitting

Overview

Why Parameter Fitting?

Quick Start

Detailed API

Constructor

Adding Experimental Data

Setting Initial Guesses

Recommended Initial Guesses

Fitting Parameter Types

Configuring Pressure Search

Running the Fit

Getting Results

Predicting Onset at New Conditions

Algorithm Details

Levenberg-Marquardt Optimization

Convergence Criteria

Onset Pressure Calculation

Best Practices

Data Quality

Initial Guesses

Validation

Troubleshooting

Fitting Fails to Converge

Fitted Parameters Outside Physical Range

Poor Prediction Accuracy

Example: Complete Fitting Workflow

See Also

References