Click Image to Enlarge

Head of Plant Projects at Nakskov, Anders Jørgensen-Juu, infront of the evaporator station.

With 4,000 employees and 16 production sites globally (including three in Australia), the company is building a unified strategy to achieve maintenance excellence—combining predictive technologies, structured planning, and mobile tools to maximize uptime and asset performance.

At the centre of this change is a smart approach toward predictive maintenance, exemplified by the work done at Nakskov around a notoriously unreliable piece of equipment: the steam dryer. Modern dryers such as the one at Naskov use hot air from primary steam to dry pulp more efficiently. They also recover and reuse secondary steam, making the process more energy-efficient and sustainable.

“Maintaining the pressure inside the dryer is crucial,” explains Head of Plant Projects at Nakskov, Anders Jørgensen-Juul. “We typically have about 2,5 to 3 bars inside, while the ambient pressure outside is much lower. Our challenge was with the outlet valve, which suffered from multiple breakdowns.”

To address this challenge, Nordic Sugar implemented a predictive maintenance system at its Nakskov plant powered by machine learning. By collecting and analyzing sensor and historical failure data, they trained a model to predict component breakdowns and estimate the remaining useful life (RUL) of parts.

Click Image to Enlarge

Dashboard for monitoring RUL development for the outlet rotary valve for the steamdrier.

According to Jørgensen-Juul, the initial results were encouraging—predictions were accurate within 13 days in the first year, enabling smarter maintenance scheduling and reducing unnecessary part replacements. Challenges arose in 2023 when needed but unplanned equipment modifications affected model accuracy, highlighting the importance of system stability and data consistency.

“Despite setbacks, the initiative has proven valuable both operationally and environmentally. Predictive maintenance has given more insight for future possibilities to extend equipment life, minimize downtime, and align with Nordzucker’s sustainability goals,” continues Jørgensen-Juul. Going forward, the company plans to cautiously expand machine learning use for high-impact components.

From Reactive to Predictive: Why the Steam Dryer Was the Perfect Test Case

“The steam dryer in Nakskov had a very unpredictable failure pattern, disturbing our seasonal production cycles,” explains Jørgensen-Juul.

“Existing diagnostic tools were not accurate enough, and this made it the perfect case to test predictive maintenance driven by machine learning.”

Click Image to Enlarge

Outlet rotary valve for the steamdrier.

The plant’s challenge was clear: break free from unplanned shutdowns and leverage data to predict failures before they occur. However, integrating machine learning into their distributed control system (DCS) and existing sensor infrastructure revealed a key insight early on.

“We assumed we had 'big data' from years of collection, but in reality, very little of it was usable. This forced us to shift our mindset from ‘more data’ to ‘right data.’ Now, we’re strict about what we collect and why.”

Building the Business Case: Scheduling, RUL, and Seasonal Strategy

For a seasonal industry where sugar beet campaigns only run for four months per year, timing is everything, notes Jørgensen-Juul. The goal of predictive maintenance wasn’t just to avoid unexpected breakdowns—it was to optimize inspections and bundle repairs during the narrow maintenance windows during campaigns.

“Being able to estimate Remaining Useful Lifetime (RUL) is incredibly valuable to us. If we can trust a machine learning model to give an accurate RUL, we might skip unnecessary inspections altogether between seasons.”

While early machine learning trials showed promising direction, they also highlighted the complexity of modelling failure. After reaching a prediction accuracy of ±13 days over a 120-day period—short of their ±5-day goal—the hydraulic system was rebuilt, rendering the previous training data obsolete.

“We’re now back to collecting data for three more production periods before relaunching the model.”

Operator-Driven Maintenance and Mobile Digitalization

“Predictive analytics is only one part of the bigger picture,” Jørgensen-Juul says. “Nordic Sugar has restructured its entire maintenance framework.”

He adds that the company has restructured its entire approach, starting with the implementation of a unified SAP-based digital maintenance system. All 13 European factories now use SAP PM for work orders and failure tracking, which supports critical metrics like mean time between failures (MTBF) and helps prioritize tasks effectively.
Since 2022, every maintenance employee has been equipped with a smartphone running the Mobile Work Order (MWO) system by 2BM Software. This digital tool allows staff to report faults using images and receive real-time updates, improving responsiveness and clarity.

“Scheduled maintenance has also become more efficient, with around 60% of planned work now automated via SAP. This shift reduces reliance on manual memory and coordination, while also ensuring compliance with legal inspection requirements,” highlights Jørgensen-Juul.

According to Jørgensen-Juul, a notable innovation is the pilot program in Nakskov focused on operator-driven maintenance. Here, technicians who operate the equipment during production are also responsible for its upkeep. This dual role enhances their understanding of operational conditions and leads to faster, higher-quality maintenance outcomes.

To further streamline efforts, Nordic Sugar has introduced a criticality classification system for all assets. Equipment is ranked A, B, or C based on its business impact, allowing teams to allocate time and resources where they matter most, and deprioritize less critical assets.

Click Image to Enlarge

Beet reception in Nakskov.

Underlying these technical changes is a cultural transformation led by strong leadership: “Reaching 60–70% maintenance efficiency is possible with relatively little effort. But 90%?

That takes leadership, attention to detail, and a shift in workplace culture,” Jørgensen-Juul emphasizes.

Maintenance managers must clearly communicate why predictive tools are being adopted and what value they bring.

“Clear communication from leaders about the reasons for adopting predictive tools—and their value—is crucial,” he stresses.

Furthermore, maintenance staff must be trained to interpret predictive data accurately. Skilled technicians play a key role in defining true equipment failure, providing the essential feedback needed to refine and train predictive models effectively.

Scaling Up—and Knowing When Not To

Interestingly, Nakskov remains the only site currently applying machine learning to RUL estimation. The reason? “The business case only makes sense for equipment with unpredictable failures, short mean time between failures, and where low-cost methods don’t suffice.”

Machine learning projects are currently limited in number due to resource constraints and the parallel push for green transformation. “We're prioritizing cases in process optimization for now, which are more straightforward. But we believe predictive maintenance will scale across the industry as tools mature.”

For companies just starting their predictive maintenance journey, Jørgensen-Juul’s message is clear:

“Start small. Choose equipment that’s critical, already monitored, and breaks down 2–3 times per year. If you can nearly solve the issue without AI, you’ll better understand how complex data preparation really is. Don’t overreach early—build confidence first.”

Over the next five years, Jørgensen-Juul sees a split trajectory in the sugar industry: some companies will try to scale too fast and fail, while others will build step by step from early successes. The company hopes to play an active role in knowledge sharing.

“In the end, all companies benefit when we share what works. Predictive maintenance can’t be scaled alone—it takes community, leadership, and practical wisdom.”

Text Nina Garlo Photos: Nordic Sugar