High-Fidelity Airborne Ultrasound Array Simulator Realized by Physics-Informed Neural Operators

Abstract

Airborne ultrasonic arrays have been widely used in contactless suspension, object recognition, directional audio, and haptic synthesis, but the development often relies on high-cost and time-consuming simulation and experiments. To solve the problem of slow calculation speed of traditional numerical software, this study proposes for the first time a method using physics-informed neural operators to perform air-coupled ultrasonic array simulation. It not only simulates propagation of acoustic waves in a 3-D environment, but also tests whether there is a similar physical relationship with the actual platform. Compared with popular numerical simulation methods, proposed has an acceleration of approximately one order of magnitude., which can support larger scale 3-D fields. This novel platform significantly reduces development threshold and cost, achieved by enabling rapid and diversified simulations that generate high-quality data necessary to support robust models.