A digital twin approach for experimental acoustic hologram optimization

The need for the accurate generation of acoustic holograms has increased with the prevalence of the use of acoustophoresis methods such as ultrasonic haptic sensation, acoustic levitation, and displays. However, experimental results have shown that the actual acoustic field may differ from the simulated field owing to uncertainties in the transducer position, power and phase, or from nonlinearity and inhomogeneity in the field. Traditional methods for experimentally optimizing acoustic holograms require prior calibration and do not scale with the number of variables. Here, we propose a digital twin approach that combines feedback from experimental measurements (such as a microphone and an optical camera) in the physical setup with numerically obtained derivatives of the loss function, using automatic differentiation, to optimize the loss function. This approach is number of transducers times faster and more efficient than the classical finite difference approach, making it beneficial for various applications such as acoustophoretic volumetric displays, ultrasonic haptic sensations, and focused ultrasound therapy.