Machine learning methods to boost the performance of food companies

Food processing involves a series of unit operations used to produce food products that meet regulatory and quality requirements. To achieve this, process parameters are continuously adjusted to compensate for variability in raw materials, process deviations and production variabilities. In food processing, the relationships between process parameters and industrial performance indicators are not necessarily known, and several performance objectives must be optimized simultaneously.

A methodology combining machine learning modelling and multi-objective optimization has been developed to improve the performance of food companies. Machine learning methods model performance indicators based on parameters collected throughout the manufacturing process. Explanatory approaches have been implemented to understand these predictions and justify them in light of expert knowledge. These models feed into a multi-objective optimization algorithm that identifies the optimal manufacturing trajectories.

Overview of the industrial performance optimization strategy using an approach combining machine learning modeling and multi-objective optimization

This methodology provides a decision-making tool adapted to industrial constraints and offers suggestions for managing food production.

Collaborations

• Altho Brets for industrial data sharing 
• Mathieu Emily (Unité Mixte de Recherche IRMAR) for assistance with data processing

Read more

Perrignon M., Emily M., Munch M., Debuire P., Jeantet R., Croguennec T. (2026) A digital twin integrating multi-objective optimization to support fryer operators in managing potato crisps production, Journal of food engineering 406: 112800. https://doi.org/10.1016/j.jfoodeng.2025.112800

Contacts

Mélanie Munch 

Thomas Croguennec 

Romain Jeantet