Interface MLCorrectionInterface

All Known Implementing Classes:
HybridModelAdapter
public interface MLCorrectionInterface
Interface for integrating machine learning corrections with physics-based models.

This interface enables hybrid AI approaches where ML models augment first-principles physics predictions. It is designed for integration with AI-based production optimization platforms.

Typical use cases:

  • Bias correction of thermodynamic model predictions
  • Data-driven enhancement of equipment models
  • Real-time model adaptation based on production data
Version:
1.0
Author:
ESOL

  • Method Summary

    Modifier and Type
    Method
    Description
    double
    correct(double physicsPrediction, double[] features)
    Applies ML correction to a physics-based prediction.
    default double[]
    correctBatch(double[] physicsPredictions, double[][] featureMatrix)
    Applies ML correction to multiple predictions (batch mode).
    default double
    getConfidence(double[] features)
    Gets the confidence level for a prediction.
    int
    getFeatureCount()
    Gets the number of features expected by this model.
    String[]
    getFeatureNames()
    Gets the expected feature names for this ML model.
    String
    getModelVersion()
    Gets the model version or identifier.
    default double
    getUncertainty(double physicsPrediction, double[] features)
    Gets the uncertainty (standard deviation) for a prediction.
    boolean
    isReady()
    Checks if the model is ready for inference.
    void
    onModelUpdate(byte[] modelPayload)
    Updates the ML model with new weights/parameters.

  • Method Details

    • correct

      double correct(double physicsPrediction, double[] features)
      Applies ML correction to a physics-based prediction.
      Parameters:
      physicsPrediction - the prediction from the physics model
      features - input features for the ML model
      Returns:
      corrected prediction

    • correctBatch

      default double[] correctBatch(double[] physicsPredictions, double[][] featureMatrix)
      Applies ML correction to multiple predictions (batch mode).
      Parameters:
      physicsPredictions - array of physics predictions
      featureMatrix - matrix of features (one row per prediction)
      Returns:
      array of corrected predictions

    • getFeatureNames

      String[] getFeatureNames()
      Gets the expected feature names for this ML model.
      Returns:
      array of feature names in expected order

    • getFeatureCount

      int getFeatureCount()
      Gets the number of features expected by this model.
      Returns:
      number of features

    • isReady

      boolean isReady()
      Checks if the model is ready for inference.
      Returns:
      true if the model can make predictions

    • onModelUpdate

      void onModelUpdate(byte[] modelPayload)
      Updates the ML model with new weights/parameters.

      This method is called when an external AI platform pushes updated model parameters after retraining.

      Parameters:
      modelPayload - serialized model parameters

    • getModelVersion

      String getModelVersion()
      Gets the model version or identifier.
      Returns:
      model version string

    • getConfidence

      default double getConfidence(double[] features)
      Gets the confidence level for a prediction.
      Parameters:
      features - input features
      Returns:
      confidence score between 0 and 1

    • getUncertainty

      default double getUncertainty(double physicsPrediction, double[] features)
      Gets the uncertainty (standard deviation) for a prediction.
      Parameters:
      physicsPrediction - the physics prediction
      features - input features
      Returns:
      predicted standard deviation