An Integrated Approach of Mechanistic-Modeling and Machine-Learning for Thickness Optimization of Frozen...

Area of Science:

• Food Science
• Engineering
• Computational Modeling

Background:

• Mechanistic-modeling aids understanding of microwave-food interactions but is too resource-intensive for direct food design.
• Food developers need efficient tools for designing microwavable foods with uniform heating.

Purpose of the Study:

• To develop and validate an integrated approach coupling mechanistic-modeling and machine learning for efficient food product design.
• To optimize food product thickness for improved heating uniformity using this integrated approach.

Main Methods:

• Integrated mechanistic-modeling (electromagnetics, heat transfer) with a Bayesian optimization machine-learning algorithm.
• Validated the integrated approach against a parametric sweep method using mechanistic-modeling alone.
• Conducted microwave heating experiments with 3-D printed trays of varying thicknesses.

Main Results:

• The integrated approach demonstrated higher efficiency (45.9%–62.1% improvement) compared to the parametric sweep method.
• Successfully optimized the thickness of different-shaped products with varying datasets.
• Experimental validation confirmed the feasibility of the integrated approach for thickness optimization.

Conclusions:

• The developed integrated approach offers a robust and efficient solution for food product design, specifically thickness optimization.
• This method significantly improves heating uniformity in microwavable foods.
• The approach can be extended to a platform for multi-parameter optimization in complex microwavable food design.