Diesel engine development involves balancing fuel efficiency with emissions such as NOx and soot. To study this challenge, Convergent Science used its machine learning optimization tool in CONVERGE Studio together with computational fluid dynamics (CFD), which digitally simulates engine processes. The study varied fuel injection timing, injection duration, and nozzle spray angle, using CFD results to train a machine learning model that predicts engine responses.
The DiRECT optimization algorithm then searched the design space using these predictions to identify a suitable configuration. CFD was used to verify the selected design and compare its fuel efficiency, NOx, and soot results with the model predictions. The approach also indicates potential use in other CFD applications, including wind farms.
Image generated by: ChatGPT