Did You Know There’s a Delicious Blend of AI, Automation and Chocolate at Siemens?

In a Siemens development lab reminiscent of Roald Dahl’s whimsical world, AI transforms chocolate production, blending tradition with futuristic precision.

In the heart of a chocolate factory, where the aroma of cocoa fills the air, a technological revolution is taking place, thanks to artificial intelligence (AI). This isn’t a scene from a futuristic Roald Dahl novel but a real-world application demonstrated by Siemens Digital Industries. Their setup, featuring conveyor belts and chocolate bars, might seem typical at first glance. However, the process behind it is anything but ordinary, showcasing how AI can bring efficiency and precision to even the most delightful of industries.

At Siemens, the Intelligent Infeed Demonstrator machine stands out as a prime example of innovative thinking. It faces a seemingly straightforward task: arranging randomly placed chocolate bars into neatly spaced slots on a conveyor belt. Yet, achieving this consistency is far from simple. Martin Bischoff, a specialist in Virtual Mechatronics at Siemens, highlights the complexity of this challenge. The solution? A smart application of reinforcement learning, an AI technique inspired by how humans learn from trial and error.

Michel Tokic, another expert in the field, further explains that reinforcement learning allows the AI to start from scratch, making guesses and learning from feedback until it masters the task. This method mimics the natural learning process, such as a child learning to walk, gradually improving through perseverance and adaptation.

What’s truly innovative about Siemens’ approach is how they train their AI. Before letting the AI take control in the real world, it undergoes rigorous training in a virtual environment, a digital twin of the actual setup.

After about 72 hours of training with the digital twin (on a standard computer; about 24 hours on computer clusters in the cloud), the AI is ready to control the real machine. That’s definitely much faster than humans developing these control algorithms

Martin Bischoff, Siemens

This strategy ensures that any mistakes made during the learning phase don’t have costly consequences. Remarkably, the AI can be ready to manage the real machinery in as little as 24 hours of training on powerful cloud computers, a testament to its efficiency and the power of modern technology.

Siemens pathway

The success of this project is not just about getting chocolate bars to line up correctly. It represents a significant leap forward in manufacturing technology. By employing AI in such practical, everyday tasks, Siemens has demonstrated that machines can learn and adapt to complex scenarios, potentially transforming entire industries.

Beyond the technical marvel, the AI Motion Framework developed from this project signifies a broader shift. It allows machines to handle tasks that have not been explicitly programmed for them, offering flexibility and efficiency previously unattainable. Thomas Menzel from Siemens sees this as just the beginning, with the potential to revolutionize how production machines are used and controlled.

In essence, the collaboration between AI and chocolate production at Siemens is more than a novelty; it’s a glimpse into a real-world example of where automation and AI can improve production. It promises a world where the efficiency of AI can be applied to even the most nuanced tasks, making scientific industries more adaptable, efficient, and, in this case, a little sweeter.

Staff Writer

Our in-house science writing team has prepared this content specifically for Lab Horizons

Leave a Reply

Your email address will not be published. Required fields are marked *