Removing Film from Pallets with Artificial Intelligence

Artificial intelligence makes it possible to remove film from all types of pallets. For the first time in the world, Danish company VARO has built a machine that can do this, and Siemens has acted as a technology partner and provided the platform that keeps the system running.

Vision Technology and AI

For the human eye, it is the simplest thing in the world to see whether the film has been removed from a pallet or not. But for a machine, it has been nearly impossible – until now.

The Danish machine builder VARO, which supplies machines to, among others, the food and pharmaceutical industries, has developed an unwrapper - a machine that automatically removes film from pallets. The machine is now supplemented with a solution based on vision technology combined with artificial intelligence, which detects whether all the film has been removed. With this solution, the process can take place entirely without an employee needing to spend time monitoring or removing film residues.

When a pallet arrives at the factory or warehouse, the film is automatically cut off and sucked into a waste container, and regardless of the lighting conditions, the size and shape of the product, or the thickness of the film, the combination of the camera and the algorithm can figure out by itself whether all the film has been removed.

As the first in the world, we have automated a traditional manual process, and because the process is unmanned, we need to give customers process assurance. Therefore, together with Siemens, we have developed a system based on AI, where our algorithms provide certainty whether there is film left or not.

Arne Lundfold Bjerring, Technical Director at VARO

Opens New Benefits for Customers

Arne Lundfold Bjerring, Technical Director at VARO

The machine itself frees up time for employees at VARO's customers. They no longer need to spend time cutting off film, and the operator is not tied to monitoring the machine. This is a big advantage for all industrial companies where it has become harder and more expensive to find labor. At the same time, the solution helps to digitize a traditional manual process, which provides advantages for both the customers and VARO itself.

“The trend is that our customers need to continually improve their performance, while there are fewer people around the machines. Through digitalization and artificial intelligence, we can offer additional services to our customers in the form of knowledge and data collection, which they can use to improve their performance, and which we can use to build better machines,” says Arne Lundfold Bjerring.

Traditional vision technology is good for ordinary inspection and quality control, where a product must match a picture 100 percent. But it cannot be used to remove film, as the possibilities for shape, color, thickness, light, and glare create almost endless combinations.

“Deep learning technology is suitable for solving all the challenges with the many film variations. So the reason we have used artificial intelligence for the unwrapper is that we encountered a problem where a vision solution combined with deep learning was the best and only solution,” says Bjarke Holm Thomsen, who is a development engineer and the primary developer of the solution.

Bjarke Holm Thomsen, Development Engineer at VARO

From Mobile Phone to Industrial Edge

Siemens-VARO-080523-004 - 9f118c9c-509f-49db-bd5a-e8da95e4553f

The process of teaching the unwrapper to detect whether all film has been removed began with Bjarke Holm Thomsen taking lots of photos with a mobile phone. Using the photos, a regular PC, and open-source libraries like Fast.ai and PyTorch, he created a proof of concept. He could quickly see that the model could solve the core task, and he began training it with more images from real production environments. At the same time, he also discovered that VARO’s challenges lay elsewhere.

“We have a good understanding of our core task and how image analysis can help us solve it. But then you encounter issues with the entire ecosystem, such as how we monitor the solution, and how we update the models - all the tasks that support the core task,” says Bjarke Holm Petersen.

The solution to that problem came from Siemens, both in terms of hardware and support. VARO borrowed some hardware to further develop the model and were introduced to how the ecosystem could be built up in Siemens Industrial Edge.

The AI technology itself is no longer so complex, but the infrastructure around it is. So it's a great help that Siemens has invested so many resources, both in setting up the ecosystem, with the entire management component, and with concrete hardware solutions, such as how we get connection through Industrial Edge to IPCs, which is a bit different than ordinary PLCs.

Bjarke Holm Thomsen, Development Engineer at VARO

Think stable hardware – also for AI solutions

For Siemens, it makes perfect sense to invest time in helping machine builders develop solutions based on artificial intelligence.

“At Siemens, we want to establish ourselves as a partner for machine builders working with artificial intelligence. We see great potential and believe AI will become a competitive parameter for Danish machine builders. At the same time, we have good solutions that fit well with the PLC and HMI solutions we already have among machine builders,” says Søren Jakobsen, Head of Industrial AI at Siemens Denmark.

He also reminds others interested that it is important to remember the hardware and the ecosystem.

“Solutions based on artificial intelligence must of course be able to solve the task, but it is also important to think in terms of industrial grade hardware. Machines often run for 10, 15, or 20 years, and the components in them are always industrial grade. So it doesn’t work to build an AI model with a camera system or other modules that cannot withstand dust, vibrations, or large temperature fluctuations,” he says.

Søren Jakobsen, Head of Industrial AI at Siemens Denmark

Opens New Possibilities for VARO

Siemens-VARO-080523-013 - e33c29dd-8f2a-4099-a98c-7d528e65cfc1

VARO’s technical director, Arne Lundfold Bjerring, is certain that the collaboration with Siemens has accelerated the process. Especially because VARO is a relatively small company without a large team of developers for such systems. But even with their small team, they have already started to see new opportunities. The benefits for both VARO and its customers do not end with the specific solution for the unwrapper.

“We are constantly getting ideas about how we can build this as a feature into our other machines to optimize customers’ processes. The same technology, for example, can be used for canned food production, where you might only need to discard two cans instead of ten, or for carpet rolls, where it’s almost impossible for humans to inspect a five-meter-wide roll. In this way, we see it as part of the Industry 4.0 mindset, where artificial intelligence is used to continuously optimize the performance of machines,” he says.

Thus, the technology also opens new possibilities for bringing customers closer.

“We have gotten used to software for everything from phones to electric cars needing to be updated via subscriptions, so the natural next step will be for us to offer updates with ongoing improvements to our customers, along with knowledge and data collection about their machine’s performance,” says Arne Lundfold Bjerring.