Prediction of Aerodynamic Broadband Noise Generated from a Flat Plate based on Machine Learning

Abstract

We propose a methodology for the prediction of the broadband noise generated from a flat plate based on the decision tree in the machine learning. First, we developed the wind tunnel with an anechoic room consisting of the sound insulation walls and the sound absorbing materials for the purpose of its accurate measurement of the aerodynamic noise. The pressure PSD, which is to be the training data, is extracted from the measured broadband noise in the wind tunnel test, the broadband noise is predicted using the regression model. In the scope of this study, the model analysis of the pressure PSD could not predict the spectral distribution of its broadband noise with the individual function. The prediction of the machine learning could predict not only the broadband noise but its discrete frequency noise to the same levels as the actual measurement. In the prediction of the machine learning, the regression model trained on a large number of training data could predict the actual noise with the small error.