Class BatchParameterEstimator

java.lang.Object
neqsim.process.calibration.BatchParameterEstimator
All Implemented Interfaces:
Serializable
public class BatchParameterEstimator extends Object implements Serializable
Batch parameter estimator using Levenberg-Marquardt optimization.

This class provides batch (offline) parameter estimation for process models using historical or experimental data. It bridges the process calibration framework with NeqSim's existing Levenberg-Marquardt optimizer in the statistics package.

When to use BatchParameterEstimator vs EnKFParameterEstimator:

  • Multiple parameters, batch/historical data: BatchParameterEstimator (this class)
  • Live streaming data, uncertainty tracking: EnKFParameterEstimator
  • Single parameter tuning: Adjuster class

Usage Example:


// 1. Build your process
ProcessSystem process = buildProductionNetwork();

// 2. Create batch estimator
BatchParameterEstimator estimator = new BatchParameterEstimator(process);

// 3. Define parameters to estimate
estimator.addTunableParameter("Pipe1.heatTransferCoefficient", "W/(m2·K)", 1.0, 100.0, 15.0);
estimator.addTunableParameter("Pipe2.heatTransferCoefficient", "W/(m2·K)", 1.0, 100.0, 15.0);

// 4. Define measurements
estimator.addMeasuredVariable("Manifold.outletStream.temperature", "C", 0.5);

// 5. Add historical data points
for (HistoricalRecord record : historicalData) {
  Map<String, Double> conditions = new HashMap<>();
  conditions.put("feedStream.flowRate", record.getFlowRate());

  Map<String, Double> measurements = new HashMap<>();
  measurements.put("Manifold.outletStream.temperature", record.getOutletTemp());

  estimator.addDataPoint(conditions, measurements);
}

// 6. Configure and solve
estimator.setMaxIterations(100);
BatchResult result = estimator.solve();

// 7. Use results
result.printSummary();
double[] estimates = result.getEstimates();
double[] uncertainties = result.getUncertainties();
Version:
1.0
Author:
ESOL
See Also:

  • Field Details

    • serialVersionUID

      private static final long serialVersionUID
      See Also:

    • logger

      static org.apache.logging.log4j.Logger logger
      Logger object for class.

    • processSystem

      private final ProcessSystem processSystem
      The process system to calibrate.

    • tunableParameters

      private final List<BatchParameterEstimator.TunableParameter> tunableParameters
      Tunable parameters.

    • measuredVariables

      private final List<BatchParameterEstimator.MeasuredVariable> measuredVariables
      Measured variables.

    • dataPoints

      private final List<BatchParameterEstimator.DataPoint> dataPoints
      Data points for calibration.

    • maxIterations

      private int maxIterations
      Maximum number of iterations.

    • useAnalyticalJacobian

      private boolean useAnalyticalJacobian
      Whether to use analytical Jacobian from ProcessSensitivityAnalyzer.

    • function

      private ProcessSimulationFunction function
      The underlying function for the optimizer.

    • optimizer

      private LevenbergMarquardt optimizer
      The Levenberg-Marquardt optimizer.

    • lastResult

      private BatchResult lastResult
      Result from the last solve.

  • Constructor Details

    • BatchParameterEstimator

      public BatchParameterEstimator(ProcessSystem processSystem)
      Creates a new batch parameter estimator.
      Parameters:
      processSystem - the process system to calibrate

  • Method Details

    • addTunableParameter

      public BatchParameterEstimator addTunableParameter(String path, String unit, double lowerBound, double upperBound, double initialGuess)
      Adds a tunable parameter.
      Parameters:
      path - path to the parameter (e.g., "Pipe1.heatTransferCoefficient")
      unit - unit of the parameter
      lowerBound - minimum allowed value
      upperBound - maximum allowed value
      initialGuess - initial guess for optimization
      Returns:
      this estimator for chaining

    • addMeasuredVariable

      public BatchParameterEstimator addMeasuredVariable(String path, String unit, double standardDeviation)
      Adds a measured variable.
      Parameters:
      path - path to the measurement (e.g., "Manifold.outletStream.temperature")
      unit - unit of the measurement
      standardDeviation - measurement uncertainty (noise standard deviation)
      Returns:
      this estimator for chaining

    • addDataPoint

      public BatchParameterEstimator addDataPoint(Map<String,Double> conditions, Map<String,Double> measurements)
      Adds a data point for calibration.
      Parameters:
      conditions - operating conditions to apply (path -> value)
      measurements - measured values (path -> value)
      Returns:
      this estimator for chaining

    • addDataPoint

      public BatchParameterEstimator addDataPoint(Map<String,Double> measurements)
      Adds a data point with only measurements (no condition changes).
      Parameters:
      measurements - measured values (path -> value)
      Returns:
      this estimator for chaining

    • setMaxIterations

      public BatchParameterEstimator setMaxIterations(int maxIterations)
      Sets the maximum number of iterations.
      Parameters:
      maxIterations - maximum iterations
      Returns:
      this estimator for chaining

    • setUseAnalyticalJacobian

      public BatchParameterEstimator setUseAnalyticalJacobian(boolean useAnalytical)
      Enables or disables analytical Jacobian computation.

      When enabled, uses ProcessSensitivityAnalyzer for more efficient derivative computation, potentially reusing Broyden Jacobians from recycle convergence.

      Parameters:
      useAnalytical - true to use analytical Jacobian
      Returns:
      this estimator for chaining

    • solve

      public BatchResult solve()
      Runs the Levenberg-Marquardt optimization.
      Returns:
      the optimization result

    • validateConfiguration

      private void validateConfiguration()
      Validates the configuration before solving.

    • buildFunction

      private ProcessSimulationFunction buildFunction()
      Builds the ProcessSimulationFunction.
      Returns:
      the constructed ProcessSimulationFunction

    • buildSampleSet

      private SampleSet buildSampleSet()
      Builds the SampleSet for the optimizer.
      Returns:
      the constructed SampleSet

    • buildResult

      private BatchResult buildResult()
      Builds the result from the optimizer output.
      Returns:
      the constructed BatchResult

    • getChiSquareFromOptimizer

      private double getChiSquareFromOptimizer()
      Gets the chi-square value from the optimizer.
      Returns:
      the chi-square value

    • computeChiSquare

      private double computeChiSquare()
      Computes chi-square from the sample set.
      Returns:
      the computed chi-square value

    • getParameterStandardDeviations

      private double[] getParameterStandardDeviations()
      Gets parameter standard deviations from the optimizer.
      Returns:
      array of parameter standard deviations

    • getCoVarianceMatrix

      private double[][] getCoVarianceMatrix()
      Gets the covariance matrix from the optimizer.
      Returns:
      the covariance matrix

    • getCorrelationMatrix

      private double[][] getCorrelationMatrix()
      Gets the correlation matrix from the optimizer.
      Returns:
      the correlation matrix

    • computeAdditionalStatistics

      private double[] computeAdditionalStatistics()
      Computes additional statistics: [MAD, bias, R-squared].
      Returns:
      array of additional statistics

    • getLastResult

      public BatchResult getLastResult()
      Gets the result from the last solve.
      Returns:
      the last result, or null if not yet solved

    • toCalibrationResult

      public CalibrationResult toCalibrationResult()
      Converts the last result to a CalibrationResult for API compatibility.
      Returns:
      calibration result, or failure if not yet solved

    • getParameterNames

      public String[] getParameterNames()
      Gets the parameter names.
      Returns:
      array of parameter paths

    • getMeasurementNames

      public String[] getMeasurementNames()
      Gets the measurement names.
      Returns:
      array of measurement paths

    • getDataPointCount

      public int getDataPointCount()
      Gets the number of data points.
      Returns:
      number of data points

    • getProcessSystem

      public ProcessSystem getProcessSystem()
      Gets the underlying process system.
      Returns:
      the process system

    • getOptimizer

      public LevenbergMarquardt getOptimizer()
      Gets the underlying optimizer (for advanced use).
      Returns:
      the L-M optimizer, or null if not yet solved

    • clearDataPoints

      public BatchParameterEstimator clearDataPoints()
      Clears all data points.
      Returns:
      this estimator for chaining

    • reset

      public BatchParameterEstimator reset()
      Resets the estimator to initial state.
      Returns:
      this estimator for chaining

    • displayCurveFit

      public void displayCurveFit()
      Displays the curve fit using the optimizer's built-in visualization.

    • displayResult

      public void displayResult()
      Displays the result using the optimizer's built-in visualization.

    • runMonteCarloSimulation

      public void runMonteCarloSimulation(int numRuns)
      Runs Monte Carlo simulation for uncertainty analysis.
      Parameters:
      numRuns - number of Monte Carlo runs