Penalization and deep learning algorithms in Holographic Direct Sound Printing to improve print uniformity

Holographic Direct Sound Printing (HDSP) is a subclass of Direct Sound Printing (DSP) method based on on-demand polymerization induced by ultrasound waves. HDSP has the capability of printing in optically opaque material and more uniquely through optically opaque barriers. This method provides layerless and fast printing as opposed to the point-based methods. However, the HDSP is highly sensitive to the nonuniformity existing in the pressure pattern reconstructed with the conventional acoustic holography methods. This results in material accumulation and some parts in the pattern solidify faster than the rest, resulting in non-homogeneous geometry of the final printed part. We provide an effective method of mitigating this issue by optimizing the acoustic image reconstruction towards more uniform printing process. The general review and comparison of various optimization techniques is presented in terms of reconstruction quality and computation time. We have introduced a new penalization technique to improve the iterative convergence and quality of the patterned acoustic pressure and uniformity printed feature size. Furthermore, we have presented a fast and efficient deep learning-based technique that provides better uniformity and Peak-Sound-to-Noise-Ratio on the generated pattern for the printing, but also robust compared to the iterative methods. The experimental results show that the printing time is shortened as well as more uniformity is observed in the final parts due to uniform reconstructed holographic image, mitigating the problem of partially solidified parts thickening before print completion. The present paper introduces a crucial step towards applying HDSP without sacrificing the feature size and overall print quality.