menu close Teknologiens Mediehus
person Konto arrow_drop_down
If you separate a spring roll, you will find that white cabbage makes up a fairly large portion of its contents. It is therefore not surprising that 20 tonnes of large white cabbage go through Daloon’s production lines in Nyborg every day.
However, not all cabbage is the same size or completely round, which is why human hands are needed to assess, weigh, and turn the cabbage so that the stem can be cut off.
But what if a robot could grab the cabbage, regardless of its size and shape, turn it over, and place it in a fixture so the stem is pointing upwards?
On top of that, a vision system could measure the size of the stem so that exactly the right amount is cut off.
For Peter Madsen, technical manager at Daloon, there were several good reasons to bring robots into the process, because the cabbage sorting work is hard and boring:
“We both wanted to do something about our monotonous repetitive work and look at how we could increase productivity and quality in the process,” he says.
On top of that, there was a question of waste when the stem is removed—the size of cabbage varies, and with 20 tons a day, there is money to be saved if as much as possible can be utilized.
The Danish Technological Institute in Odense and robot supplier and manufacturer Technicon took on this challenge when looking at new robotics solutions for the food industry:
“One of the big challenges is developing a food grade robot. This means that it must be able to withstand being rinsed and cleaned a lot in order to live up to the standards for food production,” says Carsten Panch Isaksen, senior consultant at the Danish Technological Institute and project manager for the white cabbage robot. The project is part of the EU project agROBOfood.
In the first iteration of the robot, the cabbage was placed in a basket, and a vision system was used to turn it correctly in order to cut the stem. But in the second iteration, the turning process was integrated into the gripper. For this, they used vision with AI:
“First, we turn the cabbage head somewhat randomly. A vision system with AI has been trained to recognize the stem. It’s actually not that difficult as it can do with 80–100 images before it becomes certain of it. It may be a little slower than a human operator, but we don’t think it will be a problem overall,” Carsten Panch Isaksen says.
With the current prototype, Carsten Panch Isaksen estimates that the gripper is 80 percent complete. However, it needs to be used for a longer period of time in an industrial environment in order to examine its durability and wear:
“Maybe we need to go through a third iteration before it’s ready,” he says.
At Daloon, Peter Madsen is positive about the solution. Automation has been an intrinsic part of productivity improvements for many years, and Daloon already has 32 industrial robots—both so-called SCARA robots and six-axis robots. But he also has certain reservations:
“Such a gripper probably requires a little more development, so a third version is needed, which can show its worth in a harsh industrial environment. But right now, unfortunately, we have had to prioritize our resources differently, so it will have to wait a bit,” he says.
For Carsten Panch Isaksen, the project fortunately does not stop with Daloon. The vision system and the integrated AI can also be used elsewhere:
“We have used the same solution to look at tomatoes in a nursery, and it showed that the model behind it doesn’t need to be trained on thousands of images and can therefore be relatively easily implemented in other industries. The challenge right now is to make it more user-friendly so that the local technician or production manager can feel comfortable with such a solution,” he says.
