Jet noise prediction via low-order machine learning

The intense acoustic emissions from supersonic jets can limit aircraft operation and cause serious hearing damage. It would therefore be beneficial to design low-noise engines. However, most of the factors that control noise production are decided early in the design process when it impractical to perform the simulations or experiments necessary to characterize acoustic emissions. As such, a practical design tool must be developed so engineers have the option to consider noise in future designs. In this work, a deep neural network (DNN) approach coupled with proper orthogonal decomposition (POD) has been explored as a method of predicting jet noise. It was found that coupling POD and DNN could produce a model capable of estimating noise production to within a few dB over a broad range of operating conditions while using a minimal amount of training data.