Predicting dimensional deviation for sintering-induced shrinkage of Al2O3 ceramics manufactured by vat photopolymerization: Using convolutional neural network

Abstract

Ceramic additive manufacturing technology facilitates the rapid production of components characterized by intricate geometries. Nevertheless, the dimensional accuracy of ceramic components is often compromised due to anisotropic shrinkage occurring during the sintering process. This study introduces a dimensional deviation model to represent the size variations of cylindrical components after the sintering process. Moreover, the method is expanded to encompass annular components featuring both inner and outer contours, where contour shrinkage is subdivided into self-contour deviation and influenced deviation. Additionally, a convolutional neural network is employed to predict deviations by leveraging a novel learning strategy that utilizes the established dimensional deviation model of solid cylindrical ceramic components to acquire the unknown deviation model of annular component contours. The model has been validated using a solid cylindrical component with a radius of 10 mm, as well as an annular component with an outer contour radius of 9 mm and an inner contour radius of 4.5 mm. The verification results demonstrate that our model achieves accurate predictions of the shrinkage size for these components. The proposed model is expected to replace conventional shrinkage coefficient methods, providing more accurate and reliable predictions for practical applications.