What role does AI play in your sensor solutions and systems?

Artificial intelligence is not just a technological extension for Sick, but a central component. It is specifically integrated into sensors to make machines smarter, more flexible and more efficient. It is now impossible to imagine industrial applications without AI-supported systems. User-friendly AI tools can be trained using examples. Data collection, training and execution take place directly on the sensor. Thanks to AI, sensors learn from examples and adapt to changing conditions in real time. This opens up completely new automation horizons in industrial practice.

How do you use AI in industrial image processing?

AI opens up considerable efficiency gains in image processing. Systems such as the 'Inspector83x' show that conventional rule-based approaches are increasingly being replaced by deep learning-based methods. The sensor can be trained to differentiate between good and bad parts using just a few sample images. Our pre-installed 'Nova' software enables immediate commissioning without any prior knowledge of image processing. Even complex fields of application such as high-speed inspections, reflective components or rapidly changing product variants can be mapped flexibly with edge learning. The combination of AI and intuitive operation makes image processing easy.

Can you give a practical example of how AI improves quality assurance?

AI is already making a valuable contribution to quality assurance - across all sectors, from the automotive industry to pharmaceuticals and the food industry. Applications range from inspecting solder joints and checking for completeness to detecting anomalies in packaging.

An illustrative example is our AI object detection tool 'AI Object Detection', which is used to search for, count and locate objects, for example during completeness checks. The tool ensures that the objects are in the correct position and are present in the correct quantity and without defects before the production or packaging process continues. AI Object Detection supports on-device training and is suitable for finding features within objects in captured images. The tool is trained with up to 100 images and is therefore particularly suitable for counting applications. The result is displayed as a dot on the detected object.

What are the benefits of AI for security applications?

Safety applications benefit in particular from AI, as it is able to draw reliable conclusions from large volumes of data - in real time. For example, safe route monitoring with reliable collision avoidance with people and obstacles in the outdoor area is possible. A safety system continuously measures the distance between the mobile machine and potential obstacles in the path of travel. This allows it to react intelligently and situationally to collision hazards using active and adaptive brake support.

How does the combination of edge computing and AI work at Sick?

We were pioneers in on-device learning - an approach in which sensors train and apply AI models directly on the device. In conjunction with edge computing, AI is executed where the data is generated: on the sensor itself. This minimizes latency times, reduces data volumes and enables autonomous, high-performance decision-making - without a cloud connection. Sensors such as the 'Inspector83x' learn directly in use, under real conditions. If more accurate results are required or a large variance is added, the 'dStudio Cloud Service' can be used to optionally train AI models that further increase the inspection accuracy and speed of the sensor. Collaborative data management enables effortless handling of large training data sets and accelerates the successful implementation of large-scale projects.

What data is crucial for your AI models?

Representative, realistic data under production conditions is crucial. The sensors are trained directly with images of the objects, whereby just a few examples are often sufficient. In addition, we use data from over a decade of operational experience in complex applications. For specific requirements, the combination of edge learning and cloud tools such as dStudio offers a modular architecture that allows companies to decide for themselves which data should be processed locally and which centrally - always with a view to data protection, performance and scalability.

How do you deal with explainable AI?

For Sick, explainability is not an abstract concept, but the basis for trust and acceptance. Visual feedback, clear dashboards and browser-based user interfaces make deep learning processes comprehensible. Users can immediately see which patterns or features led to the decision. At the same time, standardized tools and modular software architectures ensure transparency, both during configuration and subsequent analysis. This not only enables a high level of acceptance on the store floor, but also rapid diagnosis in the event of an error.

How do you support customers during the introduction?

Our entry into the world of AI is deliberately kept low-threshold. User-friendly tools such as Nova or dStudio also enable SMEs to implement AI projects without external experts. At the same time, we offer comprehensive training and consulting concepts, from concept development to integration and data preparation. The important thing here is that people always remain at the center. AI should not replace, but empower. Close cooperation with the customer ensures practical solutions and sustainable implementation.

What role does AI play in the development of new products?

AI is not only changing existing products, but is increasingly shaping their development. For example, AI is being used to provide faster algorithms for bar and 2D code reading. With AI-based segmentation, only the relevant image areas are decoded in real time. This ensures up to 75 percent faster image processing and therefore higher reading speeds. The function is fully implemented in the device without the need for training in advance.

What trends do you see for AI in industrial sensor technology?

Industrial sensor technology is becoming increasingly software-defined. AI-driven systems are moving closer to the store floor, becoming more autonomous, adaptable and at the same time easier to operate. The combination of AI with technologies such as the Industrial Metaverse or Digital Twins opens up new simulation and planning possibilities. At the same time, the importance of predictive maintenance and condition monitoring is increasing - driven by adaptive algorithms. In the future, it will be less about whether AI is used and more about how it is used. Humans will remain an integral part of this - as planners, trainers and interpreters of the results.