NPPL-R Project – Gearbox sorting robot for sweet peppers
Gearbox has developed a sorting robot for sweet pointed peppers for the Harvest House cooperative, in collaboration with cooperative member Frestia. This project—part of the NPPL-R program—focuses on implementing this sorting robot in the pepper company’s production environment to further optimize its integration with the rest of the business operations. Clear quality guidelines are important for the pointed pepper sorting robot.
Sorting and packaging pointed peppers is a labor-intensive process in which packers rely heavily on human labor for quality control and for quickly switching between different packaging formats. This is largely because the product is sold in small consumer packages rather than in bulk, as is the case with block peppers. “Individually assessing and sorting pointed peppers according to different quality requirements and packaging formats makes the process very labor-intensive,” says Simone Keijzer of Gearbox. “After successfully developing a snack pepper sorting robot for and with pepper grower Frestia, the question arose whether we could do the same for pointed peppers.”
Automating manual sorting work
Frestia in Honselersdijk grows snack peppers, sweet pointed peppers, and mini truss tomatoes year-round on forty hectares, divided between the Netherlands and Portugal. All Dutch locations are connected to geothermal energy. The company is a member of Harvest House for the sale and distribution of its products. At the Honselersdijk site, products are sorted and packaged according to customer specifications. Here, pointed peppers are manually processed from 6:00 in the morning until 3:00 the following night, mostly in 200- and 500-gram packs.
Harvested pointed peppers arrive in crates at the packing facility. The crates are emptied by hand, and the peppers are sorted by quality and weight. If one of the two peppers in a pack doesn’t meet the weight requirement, another suitable pepper must be found. “Because this is a labor-intensive process and it’s becoming increasingly difficult to find good workers, we started looking for ways to automate this manual work about five years ago,” says Patrick Franken of Frestia.
Sorting robot with four lines
Franken explains: “To manage and keep the packaging process affordable, we started exploring the best way to automate it. That’s when we began discussions with Gearbox to digitize the sorting of pointed peppers by quality and weight.” Over the past three years, Gearbox’s development site has produced a sorting robot with four lines, each consisting of two robots.
The process begins with emptying the crates and singulating the peppers. Using vision technology and artificial intelligence (AI), the system determines the weight and quality of each pepper, which is then used to control the robots. The first robots in the line remove Class 2 peppers, make certain weight selections, and place the peppers onto a patented sorting disk. Franken: “The first robots can remove peppers of a certain weight—for example, if a customer doesn’t want pointed peppers weighing 90 grams in the package or wants a two-piece package totaling 230 grams.”
The second set of robots assembles the correct weight combinations. These robots can also collaborate with each other. The sorting robot can produce sets of 2, 3, 4, or 5 peppers based on weight and quality.
Guarantee of proper operation needed
The system has now been extensively tested at the development site, and it’s time for practical trials. To simulate real-world use, long-duration stress tests are needed. By running the machine at full capacity for extended periods, any remaining faults can be eliminated. The robot’s capacity is 12,000 pieces per hour.
“The stress tests are essential to know whether the machine can run for twenty hours straight, whether parts overheat, or whether there are no malfunctions,” explains the grower. These tests are being carried out in a separate hall. “The machine must integrate 100% reliably into our production process—we have no time for ‘adventures.’ With this sorting line, we can’t afford to lose a day because, for example, a piece of software isn’t working properly.”
A crucial phase has now begun to objectively demonstrate the sorting robot’s performance. Franken: “That’s why this project includes neutral technicians to validate the process. I’m not a technician myself, but I do know what I want and how it should work.” Keijzer adds: “It’s great for us that the grower’s cooperative and the grower can get extra support from Wageningen University & Research (WUR) during this crucial phase.”
Independent validation of the sorting robot
Through the NPPL-R program, WUR can independently and transparently assess the status of this technology. Analyses of the business case, product quality assessment, and technical support can help improve customer acceptance of these technologies. “We weren’t involved in the sorting robot’s development, but the parties appreciate having an independent party involved,” says Sytsma.
WUR will independently validate the sorting robot’s performance in terms of product quality compared to the current manual process. Relevant quality characteristics will be agreed upon with Gearbox, Harvest House, and Frestia, such as shape, color, weight, combined weight, product damage, and visual defects. “We will also validate the weight accuracy of the sets. Manually, the combined weight often ends up higher (overfill). If the machine reduces overfill, that’s a gain.” Keijzer expects a labor saving of 10 full-time equivalents in addition to the reduction in overfill.
Linking to the business management system
The project also involves technical support for a quality control tool. Sytsma: “We’re checking whether we can clearly define the quality requirements for both parties, which are essential for a well-functioning sorting robot.” The grower emphasizes that interpreting quality correctly is key for setting the machine parameters.
Sytsma adds: “We also want to connect machine data—which we don’t yet have—to the business management system to ensure proper integration into business and work processes. There may be bottlenecks here that we’ll need to solve.”
Solving the labor issue with robotics
Franken believes that robotization is not only a solution but an urgent necessity. “We can’t keep doing it the way we are now. The migrant labor issue plays a role here.” If the long-duration tests go well, the machine will be installed in the production environment after the end of the growing season (end of October).
Franken: “It wasn’t low-hanging fruit, as we thought—we’ve been at this for three years now. We’ve run into quite a few challenges, like figuring out how to grip a pointed pepper and how to place it down again. It’s more complex than we expected. The quality of the pointed pepper is determined from a photo, and that information must be transferred to other robots. This required adding extra cameras. We sometimes wondered, is this really going to work out? But we kept pushing forward.”
