DRONE NOISE IDENTIFICATION BY SEPARATING TONES ACCORDING TO THE LEAST SQUARES SPECTRUM AND ENERGY DISTRIBUTION

Kurzfassung

For multi-rotor drones, the dominant noise sources consist of both rotor-RPM-based tones and broadband noise. Because multi-rotor drones are relatively quiet compared to larger aircraft, their broadband noise easily blends into urban soundscapes or wind interference. Focusing on rotor-RPM-based tones (harmonic frequencies) using the least-squares spectrum method offers distinct technical advantages for reliable tracking of drones during flight. The least-squares spectrum method developed in current paper can auto-matically detect and extract tones based on their individual contributions to the overall signal. This method offers specific advantages in avoiding leakage and handling non-periodic signals. The signal pre-processing and feature extraction steps were first validated using both analytical and wind-tunnel test data obtained within the framework of the GARTEUR Action Group HC/AG-26. Two flyover test scenarios—a single-drone case and a two-drone case on a windy day—were analyzed. The flight tests were conducted at the DLR National Experimental Test Center for Unmanned Aircraft Systems in Cochstedt. Although the real flyover noise signals indicate that the harmonic tones contain only about 1.4% of the total signal ener-gy due to high background and environmental noise, the characteristics of the dominant tones were still successfully identified and extracted. Both the peak values and the peak frequencies were well captured. The fundamental blade-passing frequencies for both scenarios remained within a specific operational enve-lope. Consequently, the time-frequency tracking plot provides a definitive engineering judgement on the traceability of drones under harsh, real-world conditions.