A Learning-based Approach to Modeling and Control of Inkjet 3D Printing

This paper presents a learning-based approach to modeling and control of inkjet 3D printing. First, we propose and experimentally validate a learning-based model for inkjet 3D printing. The proposed model uses a physics-based model paradigm that has been reformulated into a neural-network-like structure. This formulation enables back-propagation and the associated benefits of data-driven model identification while retaining physical interpretation of the learned model itself. Next, we propose and demonstrate a predictive control algorithm that leverages the neural-network-like structure of the model. Back-propagation is used for efficient gradient calculations to determine optimal control inputs, namely droplet patterns for subsequent layer(s), to optimize a quadratic cost function.