Mind the gap: grinding optimized

Ensuring that each roller stand in a mill applies precisely the right amount of force needed to reduce the grain into flour is a key part of milling operations. Applying too little means the machines will not process the grain sufficiently, reducing yields and affecting product quality. Applying too much increases energy consumption and accelerates wear on the rollers.  

In roller grinding, the grinding force is controlled by adjusting the gap between the rollers.  Yet the relationship between roller gap and grinding results is notoriously complex.  Finding the optimum settings is a highly skilled task, relying on the miller’s experience and process knowledge.  

“Usually, a miller takes a sample of the product going into and coming out of each stage and assesses its temperature and granulation, either by hand or in the lab, to determine whether the machine is working as expected,” says Marcus Toma, a data scientist at Bühler. “If it does not look or feel right, or it does not pass the lab test, the miller may make adjustments.” 

The variability of milling further complicates optimum machine adjustment. Different gap settings may be required to produce different recipes of flour intended for different end uses or specific customers. In addition, factors such as product mass flow distribution and roller wear profile have to be taken into account. The output of each grinding stage can also be affected by changes in grain characteristics or variations in ambient temperature and humidity.  

From intuition to data 

In recent years, Bühler technologists have been working on ways to give millers better information about the performance of their machines. One key innovation has been the development of smart rollers with built-in temperature and vibration measurement (TVM) sensors. 

Installed inside the roller beneath its working surface, these sensors provide millers with a real-time view of conditions inside the machine. Each roller contains an array of temperature sensors along its length, allowing the miller to spot unwanted inconsistencies across the working surfaces. 

Smart rollers have been operating at customer sites for several years, giving skilled millers valuable insights into process performance. “Using the system, millers gain an understanding of the temperature profile they should expect when their machines are running well,” says Toma. “If they see a significant deviation from that profile, they use it as a cue to check the machine and potentially adjust the roller gap.” 

Marcus Toma drove the development of the Grinding Optimizer in collaboration with customers.

Learning to think like a miller

Spotting potential problems is one step on the road to improved mill performance and reliability. Fixing those problems is another. The first smart roller installations still relied on the miller’s skill and experience to make the right changes to roller gap settings. A key goal of Bühler’s SmartMill journey has been to find ways to support that decision-making process more effectively.  

That challenge has been Toma’s focus. “The relationship between gap settings, temperature profiles, and grinding performance in a roller mill is complex,” he says. “It is not easy for a computer to decide what changes to recommend, and when, to optimize the process.”  

To build a system capable of making such recommendations, Toma and his colleagues needed to create an algorithm that could think more like a miller. “We started this project in close collaboration with a major customer in Europe,” he says. “They were already operating a highly advanced mill with smart rollers.” 

Working with that customer gave the Bühler team access to rich data on the way millers operate their equipment in the real world. They could analyze historical machine data to see when millers chose to adjust gap settings, what adjustments they made, and how those changes affected machine performance. 

“We used that data to train an AI system to respond in a similar way,” says Toma. “It monitors machines in real time, identifies situations where adjusting the grinding gap is likely to improve performance, and recommends the right adjustment based on interventions that worked well in the past.” 

The partner customer has been using the new AI system, called Grinding Optimizer, since 2024. The system’s recommendations are already delivering measurable benefits. In one documented case, an AI-suggested adjustment reduced power consumption in a single roller pair by as much as 38 percent while keeping quality parameters within the target range. 

Bühler estimates that AI-supported grinding optimization can reduce energy consumption of the roller mills by 8 to 10 percent, while also improving yield and product quality. Savings in energy costs, labor costs, moisture loss reduction, and reduced roller refurbishment frequency mean that installations can deliver a positive return on investment within one to two years for most customers. 

Rolling out

With early successes demonstrating the value of the Grinding Optimizer system, Toma and his Bühler colleagues are now working to extend and improve the technology on multiple fronts. 

“One piece of early feedback from our customer was that they did not want to be overwhelmed with recommendations for small adjustments that might only deliver marginal performance improvements,” says Toma. “Millers are busy people, and they want to focus their efforts on actions that will have the biggest impact.” 

In response, the Bühler team continues to train its AI system by evaluating when millers are most likely to adopt its recommendations, then adjusting the nature and timing of future recommendations accordingly.  

Separately, the team is working to evolve the Grinding Optimizer into a full commercial product. That involves testing the system with different customers and processes, extending compatibility to include the different types of software used in mill operations, and supporting a wider range of equipment types. “We would like to make the system accessible to as many mills as possible,” says Toma. “Even if they are not operating the most advanced machines or process-control software.” 

As of 2026, several pilot mills are operating Bühler’s Grinding Optimizer system ahead of a broader product launch planned for later this year. After that, Bühler expects the smart milling journey to continue, using additional data sources, more advanced AI models, and integration across different processes in the mill to further enhance optimization potential. “Ultimately, the vision is for milling to evolve into a digitally orchestrated system that continuously adapts to raw material variability, final product specifications, and commercial objectives,” says Toma. 

BACK TO THE INSPIRATION HUB