Applications of Evolved Virtual Sensors in the Automotive Industry

Activity: Talk or presentationOral presentation


We will show two concrete successful applications of genetic programming to evolve virtual sensors. In the first application we used a symbolic regression approach to evolve virtual sensors for NOx and soot emissions of diesel engines based on data from an engine test bench. The evolved virtual sensors are highly accurate and compact and can be used to estimate emissions based solely on easily measurable engine data (e.g. RPM, fuel consumption, temperatures). In the second application we used the same approach to evolve virtual sensors for the blast furnace process for the production of molten iron. The resulting models accurately model the unobservable internal state of the blast furnace and can be used to improve the control and stability of the process.
Period4 Apr 2013
Event titleEvo* - Special Session on Technology-Transfer (EvoTransfer)
Event typeWorkshop
LocationWien, AustriaShow on map