The fast development of Urban-Air-Mobility as well as the constant growth of the air transport have made the acoustic pollution abatement a crucial requirement for the aviation industries in order to comply with the increasingly demanding constraints for the community acceptance. The aeroacoustic characterization of arrays of electrically-powered propellers is one of the most challenging issues. The vast majority of the UAM concepts under development adopt propulsion systems based on multiple propellers, for which reliable and cost-efficient aeroacoustic models are still lacking. The present paper proposes the development of surrogate models for the description of acoustic emission of multi-propeller configurations. The numerical investigation focuses on surrogate models able to take into account the effects of the propeller blade geometry (e.g., chord and twist distributions) and global propeller-array geometric parameters (e.g., propellers clearance) on acoustic performances of the whole system. An innovative Artificial Neural Network adaptive metamodelling technique is applied on a numerical database obtained through a boundary integral formulation for the solution of incompressible potential flows around lifting/thrusting bodies, followed by the application of the Farassat 1A boundary integral formulation for the noise field evaluation.