Geometric compensation of sintering deformation in binder jetting additive manufacturing based on artificial neural network

Shrinkage deformations that occur during the sintering process of Binder Jetting Additive Manufacturing (BJAM) raw blanks can reduce the dimensional accuracy of the part, structural deformation, and affect the performance and quality of the final product. Therefore, the effects of these shrinkage deformations should be considered in advance during the design and pretreatment stages of the part. Traditional methods usually involve global scaling of the part or manual local modification of the part geometry, which needs more systematicity and accuracy. They also require multiple trial and error and repeated manual adjustments during manufacturing. To improve the design efficiency and part quality, this paper proposes a method based on Artificial Neural Network (ANN). A feed-forward neural network is constructed and trained by a backpropagation algorithm. The geometric data of the sintered parts obtained from finite element analysis simulations capture the changes in the grid cell node coordinates during the sintering process. The trained network was then used to automatically compensate the part geometry to counteract the shrinkage deformation during sintering to ensure the dimensional accuracy of the final product. In this paper, two parts were compensated, raw blanks were printed and sintered using the compensated STL files, and the sintered part contours were compared for consistency with the original design CAD model. In the process, an innovative machine vision-based consistency evaluation method is proposed to quantify the compensation effect. The results show that the coincidence rate between the sintered part and the target model before and after compensation increases from 42.6 to 93.5%, implying that the dimensional accuracy of the final part is improved by 50.8%. This result verifies the effectiveness and practicality of the ANN model proposed in this paper for geometric compensation.