Modelling Relationships Between Extrusion Conditions and Quality Attributes of Expanded Snacks.

Area of Science:

• Food Science and Technology
• Chemical Engineering
• Process Modeling

Background:

• Expanded snack extrusion involves intricate relationships between raw materials, processing parameters, and final product characteristics.
• Understanding these process-structure-quality relationships is crucial for product development, optimization, and scaling up extrusion processes.
• Existing modeling approaches have limitations in fully capturing the complexity of expanded snack extrusion.

Purpose of the Study:

• To critically review and evaluate different modeling approaches for expanded snack extrusion.
• To identify the strengths and weaknesses of empirical, phenomenological, mechanistic, and machine learning models.
• To propose a practical hybrid modeling framework for improved process understanding and control.

Main Methods:

• Literature review of various modeling techniques applied to expanded snack extrusion.
• Analysis of empirical regression, response surface methodology (RSM), mixture-process designs, phenomenological, mechanistic, and machine learning models.
• Development of a proposed hybrid framework integrating controllable inputs, intermediate state variables, and quality attributes.

Main Results:

• Empirical regression and RSM are widely used due to their simplicity and efficiency.
• Mixture-process designs are suitable for simultaneous formulation and operating variable changes.
• Phenomenological and mechanistic models offer deeper physical insights, while machine learning excels with large datasets.
• A consistent finding is that operating variables influence quality via intermediate state variables.

Conclusions:

• A hybrid modeling framework, combining mixture-process models with key state variables (SME, die pressure, melt temperature), offers a balanced approach.
• This framework links controllable inputs to state variables, state variables to quality attributes, and quality attributes to product targets.
• The proposed model enhances predictive performance, physical interpretability, and industrial applicability, paving the way for digital twins and AI-driven optimization.