Speaker
Description
The rapid proliferation of Unmanned Aerial Vehicles (UAVs) demands advanced diagnostic tools to analyze rotor-generated noise. While acoustic imaging via phased arrays is widely used, standard static geometries fail to suppress spatial aliasing across varying UAV rotational speeds and complex spectral signatures. Building upon wavelet-based beamforming methodologies for high-speed rotating sources, this paper proposes a paradigm shift: a quasi-real-time variable-geometry microphone array. Instead of altering the topology class, the array dynamically morphs the mutual distance between sensors to adapt to the instantaneous noise spectrum of the target drone. This dynamic shape-shifting is driven by a stochastic Monte Carlo optimization algorithm designed for embedded electronics. To demonstrate the computational feasibility of this approach, Monte Carlo runs are benchmarked across three geometries using single and dual Intel Xeon E5-2697 V4 CPUs, and an NVIDIA Tesla P100 GPU. The results deduce the necessary TFLOPS required for edge-computing implementations. By coupling stochastic geometric adaptation with exact Doppler inversion via generalized Morse wavelets, the proposed framework drastically minimizes the Maximum Sidelobe Level (MSL), delivering an adaptive tool for next-generation computational aeroacoustics.