Artificial intelligence to optimize Industrial Cheese Production
Use case: Artificial Intelligence
How to make cheese production more sustainable and improve quality and yield
An interesting example of the application of Artificial Intelligence in the Food industry. More and more manufacturers discover how to improve their production processes by making smart use of factory data.
HAI has a lot of experience with Data Science in the Food industy. The project described below was even honored with an international award.
In this case it concerns the industrial production of cheese, but the principle of smart use of data a can be applied to many other industrial processes.
| Industry: | Food industry |
| Process: | Cheese production |
| Goal: | Optimize the production process to predictable consistent quality, maximize yield, and reduce energy consumption & waste. |
| Results: | • A.I. model to weekly calculate optimum process settings • A model that predicts the expected quality and yield for the current production run • A real-time dashboard with adjustment advice for operators. |
| Award: | Winning project of the Global SAS Hackaton for Manufacturing 2023 |
Background: Challenges in industrial cheese production
For many cheese technologists at industrial cheese factories it is a weekly recurring ritual: determining the optimal process settings for the cheese factory for the upcoming week.
This is due to the varying composition of the milk that is used to make the cheese. The composition of the milk varies depending on its source and is influenced by factors such as the seasons and specific weather conditions.
Despite these fluctuations in milk quality, the goal is always to produce cheese of consistent quality and maximize yield. This means the production process must be continuously adjusted to the quality of the incoming milk.
Therefore, many cheese manufacturers conduct a weekly analysis of current and historical data. Based on this analysis, they calculate the process settings for the production lines for the coming week to achieve optimal quality and yield.
And even during production, operators occasionally need to make adjustments to the process to ensure the best possible quality and yield.
Traditionally, craftsmanship, experience, and knowledge of the process play a crucial role here.
The Case:
Automatically Generate Insights and Predictions, and provide Recommendations for Optimal Quality, Yield, and more Sustainable operation
An international industrial cheese producer asked Team HAI/Notilyze to investigate whether it would be possible to automate the determination of optimal process settings, using Data Analysis and Artificial Intelligence.
The results were surprising: a dashboard that explores all steps of the production process and builds models to predict yield and quality. Additionally, a model was developed to advise operators in real-time how to respond to process anomalies. This enables operators in the cheese factory to adjust their production process in real-time to ensure optimum yield and quality.
An impressive example of how the smart use of data helps manufacturers maximize quality and yield.
Key success factors and a potential pitfall
For the analytical part of this project, the powerful and user-friendly analysis software from SAS Viya was used. This analysis software is available as an integrated part of the HAI Cloud Data Platform, where all data related to production and quality is structured and visualized in NextGen Digital Logsheets.
The cheese producer in question had been using the HAI Cloud Platform for years, and thus there was a wealth of structured historical information on product quality and process conditions readily available for analysis. As a result, the experts could skip the time-consuming process of collecting and organizing the data they needed, and could commence analyzing two years of historical data without delay.
For successful application of data analysis and Artificial Intelligence, it is vital to have reliable data that is structured and complete. In many cases, the process of obtaining complete, structured, and validated data proves to be very time-consuming and sometimes even impossible. If you have incomplete or unreliable data, it is impossible to perform a successfull data analysis with reliable results.
This is a pitfall that many manufacturers are not aware of: they tend to believe that with the availability of a good analytics tool and a data expert, the mission will be completed successfully. However, it often turns out that good, reliable data on quality and process conditions is not easily available.
Colin Nugteren, from Team HAI/Notilyze, explains: “That’s the power of collaboration. Without the structured historial data from the HAI Cloud Data Platform, we could never have achieved such impressive results”.
Artifical Intelligence as part of the daily operation
This example shows how the AI tool can be seamlessly integrated into the daily tasks of operators, thus becoming an integral part of their daily routine.
The integration of A.I. and analytics in the HAI NextGen Operator Logsheets enables real-time support for operators’ decision-making processes. This is how A.I. contributes on a strategic as well as a tactical level.
Award winning project in the worldwide SAS Hackaton 2023
The global SAS Hackathon focuses on innovative software solutions to address societal and business challenges. Teams comprising data scientists, technology enthusiasts, and visionaries collaborate with experts and SAS mentors to develop solutions. The Tool for Analyzing Cheese Batches was designated as the winner in the Manufacturing Domain.
Lees meer over HAI’s aanpak van Artificial Intelligence in de industrie: https://hai.nl/artificial-intelligence
Get inspired
Interested in more best practices of smart use of factory data in the Food, Chemical and pharmaceutical Industry?
Get inspired by successes of others when it comes to a Cloud Data Platform with NextGen Digital Logsheets providing real-time and historical information customized for each particular purpose: real-time logsheets for each department (operator workstations) and for each purpose (OEE, Product Quality & Food Safety, Yield calculations, KPI-calculations, Mass balance details, Positive release, Golden batch analysis, CIP-monitoring, Utility consumption details, Trending, Root Cause Analysis, Cost-price evaluation, … and much more.
Don’t miss a thing and follow us on LinkedIn:
