Machine learning-based modeling in food processing applications: State of the art.

Area of Science:

• Food Science and Technology
• Artificial Intelligence in Food Processing
• Process Optimization

Background:

• Traditional food processing relies heavily on human expertise for parameter optimization, often leading to inefficiencies in energy consumption and product quality.
• Developing data-driven approaches like machine learning (ML) is crucial for advancing food processing efficiency and product consistency.

Purpose of the Study:

• To evaluate machine learning (ML)-based approaches for optimizing various food processing operations.
• To provide a practical guide for developing and implementing ML models in food processing.
• To discuss the potential and challenges of ML, including physics-informed ML, in hybrid food processing.

Main Methods:

• Review and evaluation of ML algorithms applied to food processing operations (drying, frying, baking, canning, extrusion, encapsulation, fermentation).
• Step-by-step procedure for ML model development and practical implementation.
• Discussion on neural network training/testing challenges and limitations.

Main Results:

• Demonstrated potential of ML in predicting process kinetics across diverse food operations.
• Identified key challenges and limitations in applying ML, particularly neural networks, to food processing.
• Highlighted the promise of physics-informed ML for enhanced food processing modeling.

Conclusions:

• ML-based approaches offer significant advancements for optimizing food processing parameters, kinetics, energy consumption, and product quality.
• Effective implementation requires careful consideration of algorithm selection and understanding of model limitations.
• Future research should focus on integrating physics-informed ML for more robust and accurate food processing applications.