This is the approach being followed by STMicroelectronics and Lenord + Bauer, a specialist in motion sensor systems and integrated drive technology, in a joint project. The semiconductor specialist has developed software that converts the trained neuronal networks for its 32-bit STM32 micro controller. Using this software, the companies have developed a demonstrator for evaluating vibrations. The vibration sensor, which consists of an acceleration measuring unit and an STM32 micro controller, detects a variety of signal patterns and classifies them with the aid of the neuronal network.
The sensor manufacturer wants to implement the findings in its i3SAAC product concept. The aim is to network intelligent, integrated and interactive sensors with autonomous actuators and controllers in such a way that they evaluate the data themselves. In practice, it could go something like this: a vibration sensor mounted on the bearing of a rail vehicle is equipped with artificial intelligence and identifies damage or material fatigue at an early stage. The sensor then reports this information to the central maintenance system, which compiles and analyses the information from all sensors. The condition of the bearing can then be assessed via the representation in a vibration monitor.