Georgia Institute of Technology Materials Science and Engineering

Committee Members:

Prof. Satish Kumar, Advisor, MSE

Prof. Jonathan Colton, ME

Prof. Kyriaki Kalaitzidou, ME/MSE

Prof. Meisha Shofner, MSE

Prof. Natalie Stingelin, MSE/ChBE

Development of High-Performance Carbon Fibers: Accelerating Paradigm Shifts

Abstract:

The development of three technological pathways, to accelerate paradigm shifts in the way carbon fibers are manufactured, are studied and reported in this thesis. The first two, viz. hollow carbon fibers and small diameter carbon fibers, provide a comprehensive understanding of the process, structure, and property relationship for these continuous carbon fibers. The third pathway provides insights into the challenges and opportunities to employ machine learning models to predict carbon fiber properties by leveraging experimental data to accelerate the improvement in tensile properties in a cost-efficient manner. 

Multifilament continuous hollow carbon fiber tows with a honeycomb cross-section have been produced using a gel-spun bicomponent islands-in-a-sea precursor with polyacrylonitrile (PAN) as the sea component and polymethylmethacrylate (PMMA) as the sacrificial island component. Over 80% improvement in tensile strength has been achieved for these fibers compared to the previously reported batch processed hollow carbon fibers, along with scale-up manufacturing from single filament to 740 filament tow. The effect of precursor and carbon fiber manufacturing parameters on the structure and tensile properties of the hollow carbon fibers has been studied. Furthermore, mechanical properties of hollow carbon fiber-epoxy composites have been tested and compared with commercial aerospace grade carbon fiber composites. The effect of adhesion between the fiber and epoxy matrix, alignment of fibers in the composite along the testing direction, and various testing environments, on the composite mechanical properties has been explored. Based on their properties, and properties potential, hollow carbon fibers and their composites show great promise to replace conventional aerospace grade carbon fibers in the foreseeable future. 

Continuous multifilament carbon fiber tows with 2-3 µm effective fiber diameter have been manufactured from a PAN (island) - PMMA (sea) bicomponent precursor. These small diameter carbon fibers have tensile strength as high as 5.1 GPa and tensile modulus as high as 434 GPa in different carbonization trials. The role of smaller diameter in improving the tensile properties of these fibers is explored and the nano scale defects in these fibers have been characterized. 

Finally, the efficacy of four supervised machine learning techniques, in establishing a mathematical relationship to model the continuous stabilization and carbonization process and predicting the tensile strength and modulus of the fibers, based on the manufacturing process parameters, has been investigated. The data set consisted of 600 data points with 31 features each. The results indicate that machine learning can be used to approximate the underlying function describing the effect of the manufacturing process parameters on the carbon fiber tensile properties. 

This thesis develops a comprehensive understanding of the three technologies that each have the potential to accelerate the manufacturing of high-performance structural carbon fibers. Pursuing these studies separately or in conjunction with each other will likely bring about a paradigm shift in the way high performance carbon fibers and composites are manufactured.