Starting from a concrete use-case success factor for applying AI

“Demcon is a mechatronic system designer at its core,” says David. “We develop and produce high-quality, innovative and complex systems and products for Original Equipment Manufacturers in a wide variety of markets. The current applications of Artificial Intelligence by Demcon take place within Demcon advanced mechatronics and our own product OEMs. More than 450 colleagues work within Demcon advanced mechatronics, spread over locations in the Netherlands, Japan and Singapore. In addition, we also focus on companies such as start-ups and scale-ups that provide added value to us when it comes to AI. As Demcon, we are therefore not only a system designer, but we also want to play an active role in making companies that have potential in the field of AI successful by investing in them.” 

Starting from concrete use cases when applying Machine Learning 

“Demcon's engineering companies are aimed at a number of focus markets. It concerns High-Tech Systems & Materials, Life Science & Health, Smart Industry, Water & Maritime, Agri & Food, Aerospace, Defense & Security and Energy Systems. The activities related to Artificial Intelligence and Machine Learning cover all markets. Our general view is that it is not about AI on its own. We take system design as a starting point and combine this with our experience, customer demand and domain knowledge. We start from a concrete use case. Only then can you seriously speak of applying AI or Machine Learning, or not.” 

Strength lies in developing systems in which AI is integrated 

“Demcon generally uses AI to increase functionalities, improve performance, automate certain activities and reduce costs. From a technical point of view, we mainly focus on machine vision and time series & control. We deliver products and systems in which AI is integrated, such as wearables and robotics.” 

AI applications developed by Demcon are very diverse 

“The AI ​​applications we have developed are very diverse. They include improving respiratory equipment, automatic product classification and an autonomous inspection camera for pipelines. The latter is a Deep Learning application that is based on vision software and trained on the basis of synthetic training data. The camera recognizes where it is from images and checks the quality of the pipe internally. So we initially trained this robot with 'fake data' so that the robot starts training in the real world as an adolescent instead of a toddler." 

Valuable combination: algorithms + synthetic data 

“The fact that we can generate this synthetic training data and use it to train an algorithm is an interesting combination of competencies. This allows you to train a robot or algorithm in a fast and cheap way. For example, you can also simulate the growth of a strawberry. Or simulate the quality of products, so that rare damages or defects are also sufficiently represented in the dataset.” 

Self-learning egg grading machine 

“Another example of an application that we have developed for a customer is an egg grading machine. The machine recognizes the eggs and divides them into two classes based on images. With respect to this solution, it was important for the customer that he could easily train the system himself without having much background knowledge about Machine Learning. Our algorithms and software are designed to meet these kinds of questions.” 

Autonomous sailing inspection platform 

“A final example of what we have developed is a robust, unmanned and autonomous sailing inspection platform. In the coming years, we will take major steps with AI so that the boat can recognize more complex situations on the water. This shows the diversity of customers and illustrates why AI as a competency program covers all business units.” 

Smart algorithm that quickly detects first breath through combined data 

“At the beginning of the pandemic, Demcon developed respiratory equipment in a very short time, based on the existing competence at Demcon macawi respiratory systems. We are currently looking at applications of AI to support doctors to have a faster view of when patients start to breathe on their own. Because the doctor then receives a direct signal, the patient's stay in the ICU may be shortened. Again, we use synthetic data that we add to publicly accessible clinical data. We do this because the full data sets are not yet publicly available.” 

Collaboration with very diverse parties 

“At Demcon, we speak of AI when the system learns from its environment and may therefore behave differently without explicit programming. When it comes to AI, we work together with a number of parties that, like Demcon itself, are spread throughout the Netherlands. For example, I am represented in the NLAIC, we work closely with TU/e, TU Delft, JADS and we are looking at whether we can collaborate with the RUG and other universities and colleges.” 

When working together, start from joint use cases from industry 

“When it comes to collaboration with external parties, I think it is important that the number of initiatives remains clear and that concrete use cases are used as a starting point. That is not easy at the moment. There are many different AI initiatives, each with its own agenda. What I would really like is to put use cases side by side. This way you start from the user's question and then connect to research and not the other way around. If you start working on a use-case basis, you always know why you are doing something and the step to (fundamental) research can then be made in a better way.” 

Investing in companies with potential and attracting talent 

“As I mentioned earlier, Demcon has been focusing on investing in companies that have potential in the field of AI for some time now. In this way, we contribute to an innovative ecosystem that is working with data and AI. For the Netherlands, talent and language development are very important to make AI a success. We invest in it in this way, among other things. We also hire new talent. Because Demcon is located throughout the Netherlands, people who want to work for us can usually continue to live where they live. Finding suitable talent is a challenge. Certainly because we are also looking for experienced talents who not only have knowledge that they learned in school, but also understand customer demand. What we offer our colleagues is the opportunity to work with AI in an independent and entrepreneurial way in various projects and markets.”