Digital scoops: predictive ice cream quality modeling with machine learning

Overview

Digital Scoops is a cutting-edge hybrid digital twin platform that integrates population balance physics with machine learning to accurately predict ice cream quality. The technology simulates the dynamics of fat globules and air bubbles during the freezing process, providing insights into how processing conditions such as temperature, agitation, and ingredient composition affect microstructure evolution. This predictive capability allows for rapid virtual testing, significantly reducing R&D cycles and enhancing consumer acceptance by optimizing texture and flavor.

Technical specifications

Key features:

• Simulates fat globule destabilization and air bubble dynamics to prevent coalescence, crucial for maintaining texture quality.
• Outputs size distributions (D10, D32, D50) and volume fractions as inputs for machine learning algorithms.
• Correlates microstructure metrics with consumer sensory data to predict sensory outcomes such as softness, creaminess, and flavor release.
• Provides inverse optimization for formulation and processing recommendations, enhancing product quality while reducing physical experimentation.

Technology readiness level

Digital Scoops is currently at Technology Readiness Level 6, with a validated physics model and an integrated system ready for further testing and deployment. The platform is undergoing a phased validation process in collaboration with industry partners, ensuring robustness and demonstrating return on investment through reduced R&D cycles.