[ACCESS RESTRICTED TO THE UNIVERSITY OF MISSOURI AT REQUEST OF AUTHOR.] Learning from nature livings, especially those that can respond to the stimuli and change the shape, is attracting increasing interests in a wide variety of research fields. There is a significant need of developing synthetic materials that can mimic these living systems to show dynamic and adaptive shape-changing functions. Although various fabrication methods including molding, micro-fabrication and photolithography have been developed to fabricate the dynamic materials, they all have shown some limits. At present, 3D printing is a promising technique, which provides a cost effective, accurate and customized method to form 3D structures. The recently new emerging technique, 4D printing, which employs the 3D printing to print the active materials for dynamic 3D structures, shows a great potential for various applications such as tissue engineering, flexible electronics, and soft robotics. Despite much recent progress, this technology and its application in 3D dynamic structure fabrication is still in its infancy. My Ph.D. dissertation focuses on 4D printing of programmable polymeric materials that exhibits complex, reversible, shape transformations as well as enriching the printable material library by exploring various active materials for 4D printing technology. Chapter 1 introduces the current development of active materials and methodologies. Much attention is paid to the recent progress and its merits and demerits. Chapter 2 presents a simple and inexpensive 4D printing of waterborne polyurethane paint (PU) composites that are fabricated by mixing PU with micro-size preswollen carboxymethyl cellulose (CMC) and silicon oxide nanoparticle (NPs), respectively. Chapter 3 presents the 4D printing of a commercial polymer, SU-8, which has yet been reported in this field. The self-morphing behaviors of the printed SU-8 structures are induced by spatial control of swelling medium inside the SU-8 matrix. In Chapter 4, machine learning algorithms are applied to evaluate the shape-morphing behaviors of 4D printed objects. After the model optimization by tuning the hyperparameters the obtained machine learning models enable to accurately predict the final curvatures and curving angles of the 4D printed SU-8 structures from given input geometrical information. This initial success show that these data-driven surrogate models can well circumvent the challenge of human centered trial-and-error process in optimizing the printed structures, thereby pushing the research in 4D printing to a new height.
A deep neural network model for packing density predictions and its application in the study of 1.5 million organic molecules
