Constrained Surface Design for Projection Stereolithography

Abstract

In Constrained Surface Projection Stereolithography (CSP-SL) processes, the separation of a newly cured layer from the constrained surface is a historical technical barrier and greatly limits its printable size, process reliability and print speed. Moreover, over-large separation force leads to adhesion failures in manufacturing processes, causing broken constrained surface and part defects. One of the objectives of this thesis is to characterize the separation process and identify various factors that affect the separation process in CSP-SL systems. Mathematical models for separation forces with conventional smooth constrained surface are constructed by considering the liquid resin filling process. Effects of manufacturing factors on the separation force are revealed, including resin viscosity, separation speed, the cross section of the printing object, and the initial gap between the constrained surface and the printing surface. Additionally, various factors including constrained surface thickness, printing geometry, and the oxygen inhibition layer thickness are also experimentally studied and characterized. This lays the foundation for constrained surface design, which is another objective of this research, with the goal of separation force reduction and manufacturing capability enhancement. Based on the analytical model of separation force, two novel constrained surface designs are investigated. One design is characterized by radial microgroove textured constrained surface. The effectiveness of this microgroove surface texture on reducing separation force is validated both analytically and experimentally. Test cases show that with the help of the proposed textured surface, parts with wide solid cross sections that could not be printed using conventional methods can be manufactured successfully. The influence of the microgroove texture on the printed part surface roughness is also studied. A gray scale projection approach is proposed to address the resulted surface finish problem. The other constrained surface design is characterized by air-diffusion channels made of polydimethylsiloxane (PDMS), which allows for continuous and sufficient air permeation to enhance the oxygen inhibition in CSP-SL systems. The influences of the air-diffusion-channel design on the robustness of the constrained surface and the light transmission rate are studied. It is found that the proposed air-diffusion-channel design is effective in maintaining and enhancing the oxygen inhibition effect, and thus can increase the solid cross section size of printable parts. The above two constrained surface designs, microgroove texture and air-diffusion-channel, are found effective for reducing separation force and enlarging printable size in layer-by-layer printing process. As a step further, this thesis also investigates the combination of the microgroove texture and air-diffusion-channel, which we named it as textured constrained (TC) surface, and another air permeable surface, island constrained (IC) surface, for layerless continuous printing in CSP-SL systems. The IC surface is based on a polymerization kernel surrounded with four large oxygen diffusion windows. The diffusion of oxygen in the IC surface enables the formation of a thin uncured liquid layer for continuous printing. Experiments are conducted to validate and compare the effectiveness of IC and TC surface designs for continuous printing. In addition, the surface roughness of printed parts by both approaches is characterized and analyzed. A continuous printing speed ranging from 45 mm/h to 477 mm/h has been demonstrated with our newly designed constrained surfaces. To identify the optimum continuous printing speed, machine learning techniques for modeling and predicting the proper continuous printing speed in the CSP-SL are investigated in this thesis. A synthetic dataset is generated by physics-based simulations. An experimental dataset is constructed for training the machine learning models to find the appropriate speed range and the optimum speed. Conventional machine learning techniques including Decision Tree, Naive Bayes, K Nearest Neighbors, and Support Vector Machine (SVM), ensemble methods including Random Forest, Gradient Boosting, and Adaboosting, and the deep learning approach Siamese Network are tested and compared. Experimental results validate the effectiveness of these machine learning models and show that the Siamese Network model gives the highest accuracy on predicting the optimum continuous printing speed.