Machine learning-based quality optimisation of ceramic extrusion 3D printing deposition lines

In recent years, extrusion forming has received much attention in the preparation process of ceramics and composite materials. Current research efforts mainly focus on the preparation of various extrusion pastes and the development of new extrusion processes. However, due to factors such as poor material matching at the joints and changes in material viscosity during the extrusion process, deposition lines may be uneven and discontinuous, requiring a large number of trial-and-error experiments to optimize printing parameters. In this study, a process control strategy is proposed to improve the quality of deposition lines (under-extrusion and overextrusion).The first step is to conduct experiments to explore the relationship between printing parameters and the morphology of ceramic paste deposition lines and obtain an experimental dataset. Next, machine learning (ML) methods are used to achieve process control of paste extrusion by monitoring the width of the deposition lines. The morphology of the deposition lines is captured by a CCD camera and the width is calculated through image processing on a computer.The main tasks are divided into two parts: first, using various machine learning frameworks to classify deposition lines under different parameters to select suitable printing parameters in an open-loop manner to obtain good deposition line morphology. The second part involves using previously trained ensemble learning models to predict the width of the deposition lines and establish a process model. A Proportional Integral (PI) type closed-loop iterative learning control (ILC) compensation algorithm is constructed with initial value correction, coordinating the screw speed and printing speed to achieve uniform deposition line morphology along the printing path.The effectiveness of the ILC algorithm in controlling the uniformity of deposition line width is verified through simulation, and experiments are designed to demonstrate that good deposition line quality can be achieved when extrusion is in the dead zone stage during the printing of a serpentine structure using a Direct Ink Writing (DIW) system. Simulation and experimental results confirm the effectiveness and practicality of this method.