On identifying the deterministic components of propeller noise

This study focuses on methods of processing the noise data of propulsive propellers commonly found on ubiquitous multi-rotor flying vehicles. The noise signals recorded in experiments typically contain periodic tonal and random broadband components produced by the rotational motion of the propellers, the interaction between the propellers and turbulence, and other random factors. Factors such as unsteady rotational speed, manufacturing tolerance and flow disturbance can exist, leading to time-varying characteristics of the noise signals. In this work, we made assessments of methods to identify the deterministic components of the noise signals of propellers. Considering that the different noise patterns within two adjacent periods (due to the rotation) are similar, we applied averaging methods to remove the random components iteratively. A total of three methods were employed in this work: simple averaging, exponentially weighted moving averaging, and Kalman filter averaging. The exponentially weighted moving averaging method uses a constant weight while the weighting parameter based on the Kalman filter approach is iteratively adjusted. The methods are applied to synthesized signals and results from the computational aeroacoustic simulations, demonstrating the capabilities of removing the random components. Then, the averaging techniques are also applied to the experimental results, and some characteristics of the noise signals are identified.