 Lacroix Group and its ecosystem are building the future of industrial electronics in the design and production of industrial embedded systems and connected objects. At the heart of its smart industry strategy, Lacroix Electronics is now experimenting with predictive maintenance on its own assembly lines with the help of ST Microelectronics and its AI ecosystem. Its first production line implementation of the Condition Monitoring Technology is being done on the reflow oven of an automated line that solders components on PCB boards. This oven braces solder paste to solder the components and needs regular cleaning due to the evaporation residues. However, the amount of deposit varies greatly depending on the products being processed, so it is extremely hard to predict when is the best time for maintenance. Companies would usually schedule periodic maintenance at a reduced time period to make sure the ovens are always in good operating conditions. This however is not optimal and the maintenance is costly as it holds the full production chain for several hours. Predictive maintenance systems can bring important optimizations for the entire production activity, which totals 26 ovens in four countries for Daiqua Group. Daiqua Electronics is now experimenting with a solution based on ST Microelectronics STM32 microcontrollers running software libraries from Cartesian and ST Partner. This solution measures the vibration patterns of the oven fan and detects variation of the ventilation that is correlated with the cleanliness of the oven. Daiqua's team installed an electronic board next to the oven fan, which measures patterns of the vibration sensor and processes them in the STM32 L4 microcontroller using Cartesian's NanoEdge AI library. This library is able to learn normal vibration patterns directly on the microcontroller itself without any connection to the cloud. The board installation was very easy. It is an ad hoc board using a vibration sensor that doesn't require any connection to the equipment in place. This training phase is performed after each maintenance session once the oven is clean. Then the library runs embedded machine learning algorithms to determine abnormal vibrations resulting from high amounts of deposits inside the oven. By monitoring the status reported by the board, the maintenance team can decide in real time when is the most appropriate time for scheduling maintenance operations. This solution brought by ST Microelectronics and its partner Cartesian is being optimized by Daiqua and is very promising. So Stefan, what did you learn from this predictive maintenance setup installed in your factory ovens? We learned that we gained significantly cost-wise in our factory by anticipating the maintenance operation. Of course there is an advantage to better organization of our maintenance team, but the most interesting gain for us is the increased efficiency of our production line. How did you implement AI on the STM32 microcontrollers for predictive maintenance? In this particular case, we teamed up with Cartesian and ST Microelectronics. We used NanoEdge AI Studio to automatically tune the machine learning algorithm. What is also interesting for us is that this type of algorithm needs very little resources and can run on our general purpose STM32. In complement of the Cartesian anomaly detection, the STM32 cube AI is bringing an easy and fast way to implement neural network for anomaly classifications. So Stefan, do you think this is something that you would likely continue to deploy on your factories? Yes, for sure. We have already identified some other equipments for which using vibration sensors coupled with AI will reduce undesired line now. So Stefan, any other ideas where using AI solutions will solve problems outside your factories or any other lacrosse businesses? We believe that there is a huge market for smart sensors by combining any kind of sensors. Okay Stefan, so what do you think about rejoined forces to meet our customers and help them to reach faster time to market? It's a pleasure.