Today’s thought-leaders in the industry are focused on maintaining their competitive edge and advancing their digital transformation journey by leveraging the data acquisition solutions they have implemented in R&D departments and on manufacturing floors.
Both sources of data (experimental data and historical process data) hold a wealth of knowledge that can be exploited for a variety of innovation and quality control goals including:
- Data-driven new product development and product reformulation
- Real-time quality prediction and quality monitoring
- Automated process control
Food science professionals can use FormuSense to build AI models on experimental and/or historical process data to:
- Understand the key correlations between ingredients, mixture properties, process conditions and quality
- Predict product quality for new formulations, including the use of new ingredients or new combinations of existing ingredients (such as lower-cost or sustainable ingredients)
- Optimize the formulation and process conditions to reach a tightened quality constraint (such as product texture, color or seasoning level) or cost constraint on an existing product
- Guide plant trials towards a small number of highly valuable experiments by identifying gaps and regions of interest in the design space
FormuSense allows users formulate products with targeted quality properties and minimized cost by simultaneously optimizing the selection of ingredients, the ratios in which they are combined and the process conditions under which they are combined.
Below, three successful applications of AI in food science and manufacturing, using the ProSensus FormuSense approach, are summarized.
Sustainable Snack Foods
Background
A global producer of snack foods wanted to start using readily available, locally-sourced flour types at each manufacturing facility in place of a single flour type.
Objective
Reformulate an existing product using new ingredient(s) while maintaining product quality.
FormuSense Results
Assessed the suitability of the experimental dataset to advanced modeling and developed a preliminary predictive model. Used the developed model to identify key ingredient properties that positively impact product quality (flavour and texture).
Next Steps
Perform additional experiments to further characterize the design space with better quantification of the quality properties and increased process variation.
Healthier Muffins
Background
PepsiCo wanted to make their frozen muffin batters healthier, without sacrificing taste and without introducing new ingredients.
Objective
Reformulate an existing product to achieve new quality targets, using existing ingredients.
FormuSense Results
Assessed the dataset and located deficiencies. A useful predictive model was still developed, from which, ingredients that had the most correlation with quality were identified. The existing product was reformulated with a 50% reduction in the key quality attribute.
Next Steps
Quantify missing sensory quality properties, and perform more experiments to fill gaps identified in the latent variable design space. Update the model.
Learn More
Find more details about the PepsiCo muffins project.
Better Understood Baking
Background
A baked goods manufacturer wanted to better understand correlations to product quality, and reduce plant trials for testing new formulation recipes.
Objective
Build a predictive model between formulation recipe, baking conditions and quality.
FormuSense Results
Pre-processed the dataset (including time alignment), and built a predictive model from data across 3 manufacturing plants. Performed model validation through physical experimentation. Expanded the data-driven approach to a second baked-goods product type.
Next Steps
Further characterize the design space by increasing variability in the recipe and process, expand the framework to additional baked-goods products.
Start Building AI Models for Foods
Learn more about FormuSense, and start building AI models on your foods R&D data with a FormuSense trial today.
Need more information? Request an introductory meeting.
