Sensors are technical components that convert physical or chemical quantities into electrical signals. With the advancing networking of people and things, sensors are gaining more importance. Therefore, also the number of sensors in cars experienced a significant increase. Sometimes those physical sensors are hard to install or prove to be very expensive. Here, virtual sensors can come into the game for different applications. Virtual sensors use the information available from other measurements to deduce physical quantities of interest. The advantage is that they are often cheaper, do not wear and tear, do not take up space, and often have less limitation in terms of environmental conditions. Compredict is a German startup, based in Darmstadt and specialized in AI based virtual sensors.
AVL Creators Expedition got to know Compredict at the IAA in Frankfurt and was keen on getting to know more about them. The founders Rafael Fietzek and Stéphane Foulard told us at our first meeting that their sensors use already existing signals in vehicles to measure loads like forces, torques, stresses, temperatures, or other physical quantities. These loads wear and damage components like powertrain parts, chassis parts, brake systems, tires, the steering assembly or the HV-batteries of electric vehicles. Based on the loads measured by the virtual sensors, Compredict solutions can calculate precise usage profiles for the main components of each individual vehicle of a fleet and for the exact prediction of a component’s failure.
When talking about Compredict in our Creators Expedition team we all were excited by their technology but also by their team. Of course, we always look for innovative Startups that don´t take current approaches for granted, but we also have a close look at the team! From our first contact the founders pointed out that their team consists of 23 passionate experts who have their core competencies in the fields of AI, vehicle dynamics and durability engineering.
“We are multinational, our team members have their roots in more than 10 different countries. This enables us to develop our solutions by finding a synergy emerging from a multitude of perspectives and mindsets” Co-founder, Rafael
In line with this philosophy, Compredict and AVL decided to start a collaboration project, as a fundamental step to try out its virtual sensors in the field of acoustics. The joint project with AVL focused on training AI-based virtual sensors by replacing additional hardware vibration sensors mounted at the chassis of a car to be used for acoustic investigations. The project aimed to evaluate the precision of the AI-based virtual sensors and to assess their potential to replace physical accelerometers. Furthermore, the intention was to find out which kind of acoustic events could be detected with virtual sensors and to define the focus for future applications in series vehicles or fleet vehicles based on the learning of the POC.
“Compredict‘s virtual sensor approach combined with the AVL know-how about capturing acoustic phenomena in vehicles and about objective rating of noise and vibration events offers new opportunities for bringing acoustic monitoring applications into modern series or fleet vehicles. This approach extends the possibilities for vehicle-based condition monitoring by using the information transferred with noise and vibrations. Finally, this helps customers or vehicle operators to improve and maintain the quality of their vehicles and to safe costs, moreover.” Alexander Ulz, Project Manager AVL
The collaboration project with Compredict is a good example of our Expedition approach.
“In this project we had the chance to bring our companies together, to get to know each other and to identify opportunities to gain advantages by combining our technologies. We liked the highly skilled and open minded Compredict team and we are very much looking forward to continuing or collaboration to solve exciting technical challenges together.” Alexander Ulz, Project Manager AVL
AVL and Compredict were able to join forces, so that both sides could bring in their strengths and competences.
“We learned that our technological approach is capable to represent hardware vibration sensors through virtual sensors and that the virtual sensors perform well in the required frequency range. Based on these results we also learned that there is great potential for further collaboration with AVL.” Rafael Fietzek, Co-founder Compredict