A Deep Learning-Based Real-Time Noise Prediction Of Full-Scale Helicopter Rotor

Abstract

This paper develops a Multilayer Perceptron (MLP) neural network model for predicting full-scale helicopter rotor noise. Given that the neural network model's computational speed is exponentially faster than traditional Ffowcs Williams-Hawkings (FW-H) equation-based solvers, it has the potential to be used in real-time noise prediction. This study aims to assess the capability of the neural network model towards predicting the noise of a full-scale Bo 105 helicopter rotor. The model can be used to predict noise at arbitrary spatial locations at different flight conditions within the design space. A requisite dataset of noise OASPL used to train the model is generated from a FW-H equation solver, PSU-WOPWOP. The rotor acoustic source is obtained by using a mid-fidelity comprehensive analysis code, DYMORE. It was found in the study that, the predicted noise OASPL within the design space is accurate and MLP is able to capture the mathematical relationship between input parameters.