Data-driven Learning and Modeling of Carbon Fiber Reinforced Polymer Composites PDF Download

Author: Shenli Pei
Publisher:
ISBN:
Category :
Languages : en
Pages :

Get Book

Book Description
Carbon-fiber-reinforced polymer (CFRP) composites are being widely used as lightweight and high strength material in aerospace and automotive industries, owing to their high specific modulus, high specific strength, and good corrosion and fatigue resistance. The material performance of CFRP composites highly depends on the material manufacturing process and the inherent internal microstructure. Therefore, this dissertation attempts to unveil the underlying process-structure and structure-property relations by integrating data science and informatics with microstructure characterization. Specifically, this dissertation focuses on developing analytical approaches for CFRP composites from X-ray computed tomography (XCT) images, a nondestructive testing, and three key research topics were identified. These topics include a) developing a 3D microstructure characterization approach for non-uniformly oriented CFRP composites, b) establishing physics-based features to quantify the spatiotemporal progression of tensile fractures in CFRP composites, and c) comprehending the process-structure-property (P-S-P) relations of fused filament fabricated CFRP composite through developing image-based analytical methods that quantitatively examines the microstructure variations and its effect on the tensile property. For the first research topic, a 3D microstructure analysis framework was developed to quantitatively analyze fiber morphology (e.g. fiber curvature, orientation, and length distribution) for non-uniformly orientated fiber systems using micro-XCT ([mu]XCT) images. For this purpose, numerical image processing techniques and iterative local fiber-tracking approaches were developed to extract individual fibers from congested fiber systems, and statistical distribution of the fiber morphology was formulated using tensor representation. The derived statistics were integrated with the physics-based Halpin-Tsai model and laminate analogy to estimate the material modulus. The fidelity of the characterization was validated through experimental results for injection molded short and long CFRP composites, which provided a valid alternative for finite element analysis. For the second research topic, the spatiotemporal characterization of the fracture behavior of CFRP composites was established through the implementation of in-situ [mu]XCT. The fracture features were automatically extracted from the 3D [mu]XCT image using the image processing techniques, and physics-based features were developed to quantitatively measure the progression of failure behavior. The proposed characterization approach was implemented on sheet molding compound and injection molded CFRP composites, where the spatiotemporal characterization of fracture behavior was quantified and visualized. It provided insights into the microscale failure mechanism, and the validity of the proposed characterization approach was confirmed by the strain field calculation using a volumetric digital image correlation. For the third research topic, a P-S-P approach was proposed to unveil the underlying relation between the process parameter of fused filament fabrication (FFF) and uncertainties in the microstructure of the printed CFRP composite. An image-based statistical analysis was developed to formulate a stochastic model for the microstructure distribution (i.e., fiber and void volume fraction), and analysis of variance was implemented to establish the correlation between the process parameters and resulting microstructure. The structure-property relation was investigated by employing the physics-based Halpin-Tsai model to predict the material modulus. A data-driven optimization scheme was developed for the Halpin-Tsai model to account for the complex effect from the FFF process and craze nucleation from voids; therefore, the optimized model provided an accurate estimation of longitudinal modulus of FFF parts. Further, a Monte-Carlo sampling method was adopted to investigate the propagated uncertainties in the structure-property relation.