"Tiny ML is key for SMEs to enable energy transition"

Last year, Techport and Hilton Foods Holland successfully completed the first use case Tiny Machine Learning. With Tiny Machine Learning, energy consumption and the production process can be mapped, resulting in energy savings of up to 30%. Jeroen Reiber of Hilton Foods talks about the successful collaboration as part of Techport's 10th anniversary.

    

Combating food waste
Jeroen: "About two years ago I came into contact with André Gerver of Techport through a colleague. We share the same passion for innovation. Soon we came up with the idea of setting up a use case with Tiny ML together with two students. This technology allows you to have predictions performed on the device itself (on the edge). It is safe, cheaper and faster(er) than current solutions where you first have to invest a long period of time in collecting data and storing it in the cloud. Only after a long period of time can you then draw conclusions from that data. With Tiny ML, this process is much faster because no data transfer is required. We researched one of our production machines in this way to predict maintenance to prevent failure or downtime, which in turn reduces food waste. One student demonstrated the technical feasibility. The other student worked out the strategic implication of Smart Industry for the entire organization and the business case."

Develop from the user's needs
"As a true techie, I want the basics to be in order, but also look at how digitization can help our company, our employees and our customers further. We supply meat to the Netherlands' largest grocer. Every year we want to do it a bit faster, better and cheaper. Growing together and getting better, that's the goal." Jeroen sees Steve Jobs as his great inspiration: "Jobs developed products from the needs of users, not the other way around. This is also how we apply Tiny Machine Learning. What does it benefit our employee? And our customer? And ourselves? We use data to predict maintenance, so we can prevent problems before they happen. You want to prevent machines from stalling, people from not being able to work and the meat from not being able to be processed. I find it a challenge to make all the information we get from such a machine understandable. Not only for our management, but also for the people who operate the machines, for example the operators. That they too understand what they are looking at. That way the user experience will also be better for them."

Mission: implementing Tiny ML within all of Hilton Foods
Jeroen's mission is to implement Tiny Machine Learning throughout Hilton Foods. The initial results were convincing, giving the board the green light for the next step to monitor an entire production line in the same way. "You can also use this technology within other parts of your company. For example, for energy savings or mapping the entire process from supplier to customer. Throughout the entire journey there are data points that provide interesting information to make certain processes more efficient."

Key Technology
Tiny Machine Learning is still relatively unknown, but offers a lot of potential for companies. Especially for small and medium-sized businesses because it is affordable. Jeroen is enthusiastic about the cooperation with Techport: "They have been extremely good in thinking with us about setting up the research and working out the business case. It is an accessible way to innovate without large investments or risks. You can get started with just a few hundred euros in hardware costs. At the moment, there is too little (correct) data and the quality is insufficient and the solutions are too expensive to make real strides with predictive maintenance. Tiny ML is going to enable and accelerate the successful deployment of predictive maintenance. It is rightly seen as a key technology."

Who is Jeroen Reiber? Name: Jeroen Reiber (age 53) Works as: Manager Engineering and Technical Support At: Hilton Foods Holland Since: 15 years (with a small foray into another employer) Was/Did before that: Through industrial refrigeration, I ended up in meat processing/packaging. Had my own company providing technical service to industrial butchers of various retailers in North Holland. Connection to the IJmond region: I once had an internship at Tata (then Hoogovens). And my mother lived in a cottage on the stand of IJmuiden. Wish for the next ten years 'Together towards the first green industrial zone of NL: That in four years we have a strong vision of how to make the best use of AI and Tiny ML. And that we already have quite a few practical examples of this running.