AI-Enabled Predictive Modeling of Flour Rheological Properties Using Machine Learning Techniques

Flour rheological properties such as water absorption, dough development time, stability, and extensibility are

fundamental to assessing dough behavior and bread-making quality. Conventional rheological instruments like the

farinograph and extensograph are accurate but costly and time-consuming. Artificial intelligence (AI) and machine

learning (ML) provide modern, data-driven solutions for rapid and accurate prediction of rheological properties. This

paper presents an AI-based framework integrating artificial neural networks (ANN), convolutional neural networks

(CNN), and explainable AI (XAI) methods to model rheological attributes using physicochemical, near-infrared (NIR),

and imaging data. The ANN achieved R^2 values exceeding 0.78 for farinograph parameters, while CNN models

processing 3D imaging data achieved R^2 near 0.90 for extensibility and toughness. SHAP analysis identified protein

content and specific NIR wavelengths as dominant features. The study demonstrates that advanced AI algorithms can

replace traditional rheological testing with reliable, cost-effective, and explainable predictive systems.