AI Optimization in Protein Isolate Production
ASNS Ingredient is implementing a research project under the Recovery Fund titled “Latvian Food Competence Center – Digitalization”, in accordance with the “Latvian Recovery and Resilience Facility Plan, Reform and Investment Direction 2.2 ‘Digital Transformation and Innovation of Enterprises’, Investment 2.2.1.3.i ‘Support for the Introduction of New Products and Services in Business'” implementation regulations.
Research title: “Application of Optimization Algorithms in the Production Process and Recipe Development of Protein Isolate.”
The research project is being carried out within the framework of the project of SIA “Latvijas pārtikas kompetences centrs“, No. 2.2.1.3.i.0/1/24/A/CFLA/002, in cooperation with SIA FIDEA.
Project implementation period: 02.01.2025 – 28.12.2026.
The goal of the project is to develop a prototype of an AI algorithm that will improve the recipes of produced protein isolate and by-products, increasing production efficiency and the quality of the final product.
The main activities include collecting data from production equipment, developing and validating AI algorithms, and integrating them into a cloud system. As a result of the project, a modern AI solution platform will be created, enabling faster and more accurate optimization of production processes, providing the company with a significant competitive advantage.
Within the framework of the research project, it is planned to develop artificial intelligence and linear mathematics algorithms for recipe optimization, as well as interactive panels for training and visualizing these algorithms. In addition, further development will involve analyzing model quality and exploring potential applications in the high-value food industry. The project is also expected to have a significant positive environmental impact (a green product), replacing meat and other products in food.
Planned project activities:
- Collection of laboratory data
- Development of a processing model
- Model validation in the laboratory
- Method development for implementation in the production facility
- Initial data collection from industrial equipment
- Transmission of industrial equipment data to the cloud
- Development of an administration panel
Expected result: A created and practically validated prototype.
It is expected that the implementation of such algorithms in production will reduce the number and scale of labor, time, costs, and unsuccessful attempts when trying to identify a new protein isolate recipe with specific properties. For example, if customers seek a protein isolate powder that produces a gel-like final product when used in food development, using algorithmic optimization tools, the manufacturer will be able to reach the desired result much faster and with fewer resources.
By using models capable of processing big data, the manufacturer positions itself among the few companies ready for future opportunities. Additionally, AI algorithms can help identify inefficiencies and anomalies in production processes that would not be possible without such models. It may even become possible to predict food quality non-conformities before the full production process is completed. All this will enable the manufacturer to save resources, make the company more sustainable, and improve profitability, supporting the Latvian economy.
INTERIM RESULT 02.01.2025 – 30.05.2025
Implemented activities:
Laboratory-collected data and production processes were identified and consolidated. Process descriptions were reviewed to become familiar with the production equipment and process sequence, as well as to understand the underlying chemical processes.
Code for data processing and analysis was developed using Python.
Possible data variations were identified, along with their expected impact on the results.
Technological and other constraints were identified.
Achieved result:
Available datasets were consolidated, and the production and data acquisition processes/methods were identified. The expected impact of parameters on the results and their limitations were clearly defined.
Published: 01.06.2025.
INTERIM RESULT 01.06.2025 – 31.12.2025
Implemented activities:
A review of scientific publications and other information sources was conducted.
Algorithmic and machine learning methods previously applied in the industry to solve similar problems were identified.
From the available datasets, subsets intended for testing and validation of the study were selected.
An environment for algorithm testing and result visualization was established.
Achieved result:
The most suitable algorithms and methods were identified. Core algorithms were selected for further modeling. Datasets and a prototyping environment for testing the mathematical model and presenting results were prepared.
Published: 15.12.2025.
Ziņojums latviski pieejams šeit
