The potato harvest can be quite rough. While farmers, their families and helpers used to collect the tender tubers in the field by hand, nowadays the harvester is in use. If something goes wrong here, the potatoes rumble hard over the vibrating belts, hit stones and bounce against each other on the conveyor belts. This gives the sensitive tubers 'bruises'. This can damage them so badly that harvest losses are the result.

"If the farmer knows that the potatoes are taking too many hits, he can react. For example, he can adjust the speed of the harvesting machine or take more soil onto the conveyor belt," explains Dr. Sabine Janzen, research assistant to Professor Wolfgang Maaß. Maaß and his team have set themselves the goal of ensuring greater transparency around the tuber in the future. The decisive factor here is that the farmer must first receive the relevant information - and in good time, i.e. during the harvest. With their forecasts, the researchers want to provide decision-making support right from the field. They network a wealth of information - from agricultural machinery data and price forecasts to the number and severity of impacts the potato takes on the conveyor belt.

Hannah Stein and Mirco Pyrtek work in Prof. Wolfgang Maaß's team. In order to record how many impacts the tuber takes on the conveyor belt, they have a pain-sensitive artificial tuber (on the table) harvested with it.

© Saarland University

To this end, the business IT specialists are creating a 'digital administration shell' - a kind of logbook that contains as much information as possible about the potato batch. In this context, the researchers also determine how many blows the potatoes receive - using a 'pain-sensitive' artificial tuber. The potato harvester also harvests this so-called nPotato, which means that the nPotato takes the same route through the harvesting machinery as its real counterpart. It uses sensors to detect impacts and rotations. If it gets too much, it issues a warning. Sabine Janzen explains: "In order to network physical and virtual objects, classify the impacts and draw conclusions, we combine machine learning methods, known as deep learning, with information and communication technologies."

In this way, the researchers link a large amount of data: When was the potato of which variety harvested where? What is its water content? What is it suitable for? They integrate price and financial forecasts and, in future, historical data, harvest logistics processes from the previous year, weather forecasts and, last but not least, the farmer's expert knowledge. In this way, they create a service platform for everyone involved with the tuber: From helping the farmer decide when best to bring the potatoes to market, to quality levels, whether the batch is suitable for the star chef or more for starch flour. Production machines could also be adapted on this basis so that they peel the potato skin deeper, for example. Last but not least, commodity investors can secure their purchases with appropriate quality seals. "If we look at the data from several farmers together, we can make even more far-reaching predictions. For example, chip manufacturers could decide to use a different potato variety if the harvest quality is likely to be such that there will be a problem with the other variety in three months' time," says Professor Maaß, looking to the future. In order to further develop the solution for practical use, he and his team are currently also looking for other interested partners.

Availability is sold

Researchers at TU Kaiserslautern (TUK) are working with industrial partners to develop new business models for the agricultural machinery sector: the tractor or combine harvester is no longer sold as a product, but as a service.

Many agricultural machines represent a huge investment for farmers. Yet they often only need the machines for a few days a year. Against this backdrop, the new business models that the researchers from Kaiserslautern are currently working on making commercially viable aim to 'only' sell the availability of the required agricultural machinery for a certain period of time. For the supplier, this means that they have to guarantee that the desired machine is available to their customer as close to 100% of the time booked as possible.

Thomas Eickhoff (left) and Hristo Apostolov from the TU Kaiserslautern are working on a technical system that detects machine failure at an early stage.

© TUK/Thomas Koziel

The researchers at TUK are working together with IT and telecommunications companies, software system providers, consulting firms, industrial suppliers and agricultural machinery manufacturers John Deere and Grimme, as well as drive technology company Lenze, who are providing them with the relevant machines, equipment and data for their development activities, to guarantee such availability from a technical perspective.

"In the project, we are developing an overall system that monitors the machines in such a way that we can recognize early on when a failure may occur," says Thomas Eickhoff from the Chair of Virtual Product Development at TUK. Sensors are used to provide the engineers with data on the condition of the machines. In a potato harvesting machine, for example, they can monitor the conveyor belt and collect data. "We evaluate this data in order to predict belt faults and failures in good time," continues Apostolov. The manufacturer could then, for example, arrange for a service technician to travel to the customer before the machine breaks down.

There are countless variations of tractors and other machines, and they can be equipped with different accessories depending on the farmer's needs. "If a breakdown occurs, a replacement has to be found quickly. However, this is only possible if you know exactly which part is installed where," says Eickhoff. With the digital twin of agricultural machinery, which the Kaiserslautern researchers are also developing, all the necessary data, from individual components to repair instructions, will be stored digitally in a database. To this end, they are developing an intelligent information management system in which all important information about the machines is compiled in a user-friendly way.
The current work is part of the joint research project 'InnoServPro' (Innovative service products for individualized, availability-oriented business models for capital goods), which is funded by the Federal Ministry of Education and Research (BMBF) as part of the research programme "Innovations for the work of tomorrow - research for production and services of the future".