AI in Closed-loop Process Control Applications

For years, Process manufacturers have used various tools to optimize industrial processes for specific economic objectives. Multi-variable process control (MPC), or better known as Advanced Process Control (APC), has been the most common tool to deliver benefits and increase profits. However, as advances continue on the computing technology front, manufacturers move to simplify systems and reduce reliance on complex, time-consuming modelling techniques. As such, one of their most pressing challenges lies in efficiently deploying and sustaining assets with knowledgeable and competent resources.

Thanks to robust, secure cloud services unleashing Artificial Intelligence (AI) and statistical methods, on time-series data common to the shop floor is easier. In recent years, AI has proven to be a valuable off-line analysis tool for process engineers. Still, the question has always been – when will industrial users be able to leverage AI to close the loop and remove the human from the decision process? 

In a recent briefing with Intelecy we learned of use cases where process manufacturers have successfully deployed an AI solution to perform closed-loop optimization in real-time. According to Intelecy, what was once a complex solution reserved for data scientists, custom integrations, and data lakes, has now moved to a simple solution delivered via cloud services, software-as-a-service (SaaS) and the added benefit of MLOps (Machine learning operations) to care for the models.

Intelecy, headquartered in Norway, was founded with a simple yet powerful vision - to empower sustainable industrial production by harnessing the vast amounts of data generated and utilizing it to its full potential. Intelecy’s mission is to empower industrial organizations to improve processes and reduce waste, emissions, energy consumption, and costs. The company boasts their no-code Industrial AI solution can create Machine Learning models in minutes for real-time predictions that drive efficiency, quality, and sustainability optimization.

We learned one interesting Intelecy use case - a food & beverage company based in Norway, TINE Jaeren, started production in 2014 with a facility streaming 40,000 tags continuously, representing over 250 million rows of data daily. The company aimed to optimize the yield of protein powder production, reduce variation, and be on target for quality. They already had NIR sensors measuring protein continuously, and process changes were performed manually. Using the Intelecy AI tool, TINE Jaeren process engineers built machine learning models that forecast how the protein content will evolve up to one hour into the future. Intelecy’s Gateway is streaming bi-directional OPC-UA data in real-time. Predictions from the (AI/ML) forecast-models were sent back to plant SCADA controlling the process. As a result of the project, Operators could make automatic filtering loop adjustments based on a forecast, meaning there is less variation and better yield.