Numerical investigation on Al2O3 droplet spreading and its prediction model exploration based on Harris Hawks optimization-generalized regression neural network in stereolithography

Abstract

The laying process is crucial in using stereolithography (SL) for molding Al2O3 parts. However, most studies focus on the laying process of macroscopic slurry; there needs to be more focus on microscopic exploration. Studying from a microscopic perspective can help us understand the influence of its parameters on droplet spreading and infer the macroscopic changes of the slurry based on the changes in droplet spreading to understand why parameters cause macroscopic changes in the slurry. A pseudopotential model based on Sisko's non-Newtonian behavior in lattice Boltzmann method is proposed to study the spreading process of droplets and validated using wetting characteristics. The previous layers of the platform and the printed solid are investigated to understand the effect of laying velocity on the spreading diameter, the thickness, and the both-sided contact angles. The results indicate that a higher laying velocity leads to a larger spreading diameter, a smaller spreading thickness, and a smaller left contact angle. However, it also increases the contact angle difference between the two sides, leading to uneven slurry. The droplet spreads more unevenly when the previous laying surface is the printed solid. At the same velocity, the droplet spreads with a smaller diameter, thicker thickness, and larger contact angle on the printed solid surface. Therefore, a higher laying velocity in the SL laying process is not recommended, especially when the front layer is a printed solid. Although a higher laying velocity will increase the laying area and reduce laying time, it will cause protrusions at the front edge, and inconsistent laying thickness of the same layer will affect the following photosensitive curing process. The Harris Hawks optimization-generalized regression neural network algorithm is proposed and compared with other common artificial intelligence algorithms to predict the spreading parameters. The comparison shows that the proposed algorithm provides a more stable and accurate prediction of spreading parameters.