Applying AI should be the finishing touch

NIRAS’s experts Erik Snijders, Dirk Boumans and Chris Bullock discuss the role of data, expectations regarding AI and new regulations impacting current investments for packaging lines within the Food and Beverage industry.

The role of data in packaging lines is growing fast. Where once an engineer focused mainly on mechanical aspects, now it’s about algorithms, sensors and software. “The shift toward smarter systems is irreversible,” says Erik Snijders, Senior Project Manager and team lead Packaging Engineering in The Netherlands. “We’re all generating more and more data, and that gives us a certain control we didn’t have before.”

He explains that packaging lines used to be largely mechanical systems that had to run as efficiently as possible. “Now you see self-learning systems: a sorting machine can adapt its own algorithm based on data and continuously improve its output.” And things are changing beyond the production line as well. In a warehouse, temperature measurements can automatically decide which raw materials should be used first. “You’re no longer just looking at order, but at actual quality,” he says.

“The shift toward smarter systems is irreversible,”
Erik Snijders, Senior Project Manager and team lead Packaging Engineering in The Netherlands

Unknown fuels reluctance

Not every production company embraces digitisation at the same pace. “Unknown breeds reluctance,” says Snijders. “Some machinery builders, suppliers or food producers have worked with the same technology for years, and find the transition to something abstract like ‘data’ difficult. With a mechanical line you see movement, you can repair something. In a digital environment you have to trust information you can’t physically hold.”

For Expertise Director Digitalisation, Dirk Boumans, there is also the investment aspect. “It takes budget and time. Over the long term it pays off, but that’s not always obvious immediately.”

Both experts noticed that the hesitation isn’t with the machinery builder, but with the food producer. They must invest in new systems and train people. The gap is big: some already use a MES-system to manage production, others are still hesitant. As these technologies become more common and move into day to day operations, the impact is becoming clearer from an engineering perspective too.

Chris Bullock, Engineering Director for NIRAS UK commented on these insights, “For our client’s product quality is paramount, this needs to be monitored and controlled regularly and it is usually requires a manual intervention. This does nudge manufacturers to look into new and innovative solutions.”

Quality and safety

Data raise new questions. “You see that machine builders sometimes need access to the control system and the data from a line to improve algorithms,” says Snijders. “Then you need to decide whether you want that, and what it means for IT security.”

Another issue is the quality of the data available. “With machine learning, or artificial intelligence, what you put in is crucial for what you get out,” says Boumans. “Not everyone works with sufficiently high-quality data sets.” Data are increasingly collected in real time, allowing remote monitoring of maintenance and performance. “Maintenance can even be outsourced,” adds Snijders. “The supplier can follow along and intervene before something fails.”

Dirk Boumans on working with simulation models: “This way you can see what happens if you adjust something in the process, without having to rebuild machines or entire lines.”

Services becoming more digital

The way production facilities or installations are being designed and reviewed is changing too. Besides digital 2D layouts, 3D, BIM or VR models, NIRAS is now more often working with simulation models that allow testing data, scenarios or proposed changes in a virtual factory. “This way you can see what happens if you adjust something in the process ,without having to rebuild machines or entire lines,” says Boumans. “Our services have become much more digital as a result.”

Testing designs or full master plans digitally is a fundamental shift. It also helps to look further ahead, or to justify investments that might only pay off after 3 to 5 years.

Of course, digital development comes with conditions: “You need a data infrastructure, sensors and a clear use case. Think visual inspection, predictive maintenance or order optimisation,” says Boumans.

Legislation adds uncertainty

New regulations, such as PPWR, introduce additional dynamics. Especially uncertainty about exceptions makes decision-making hard. “Clients now want to invest, but they don’t know for sure what will still be allowed in the future,” says Snijders. “Look at PET punnets for certain fruits. They’re currently on the exception list, but what if they’re removed later? Then you’d have to switch to cardboard with a plastic liner. That means major adjustments.”

For that reason companies increasingly opt for flexibility. “You build lines so they can handle different materials. That gives the freedom to respond to whatever comes. But this also demands more from machine builders, and from us as an engineering partner. We need clarity, or at least clear boundaries in requirements, before ordering machines or building packaging lines.”

AI in practice

Artificial intelligence is becoming increasingly visible in production environments, for instance through broader use of vision systems. In the fruit and vegetable sector, developments are already advanced. “With several cameras you can even measure sugar content of products and sort based on that,” Snijders explains. “That way you can send products in real time to the right customer.”

Bullock adds: “We have supported client’s looking to embrace Smart Manufacturing solutions, in one example seeking to introduce continuous monitoring and automatic optimisation of machine parameters before pack quality is decreased or maintenance is required. The result of these machine learning initiatives reduces waste, customer complaints and unplanned downtime.”

Boumans offers another example: “We have seen lines that automatically adjust their speed based on upstream or downstream data. There are systems that optimise order sequence based on pallet configuration. Fully self-learning lines are still very rare, but the building blocks are there: AI, machine learning and digital twins make it increasingly realistic.” He adds: “Applying AI should be the finishing touch of optimisation and efficiency improvement. It starts with collecting and validating data, then gaining insight, and when those steps have been taken, machine learning can be used to adjust in real time or predict when maintenance is needed.”

According to Boumans, there are also applications in production planning. “AI can help forecast when it’s best to switch, to minimise downtime.”

The message for food and beverage manufacturers is clear: the future belongs to those who turn data into a powerful tool. Mechanical reliability is no longer enough on its own. Smarter systems, enriched with real-time insight, are reshaping how quality is monitored and how production performance is assured.

Companies who embrace smart tools, from advanced vision systems to simulation models and early-stage predictive maintenance, gain far more than efficiency. They gain resilience in a market shaped by changing legislation, the demand for flexibility, and rising consumer expectations.

While investment and cultural adaptation are known hurdles, the pathway to be more data driven is proven: build a trustworthy data foundation, upskill teams and introduce smart solutions step by step.