UA cloudflare authentication

Browsing by Author "Papon, Md Easir Arafat"

Now showing 1 - 3 of 3
  • Thumbnail Image
    Item
    Additive manufacturing of two phase thermoplastic composites: a process model, microstructure and performance study
    (University of Alabama Libraries, 2019) Papon, Md Easir Arafat; Haque, Anwarul; University of Alabama Tuscaloosa
    Fused filament fabrication (FFF) based additive manufacturing (AM) of polymers and composites is a growing interest in processing tailorable parts with functional requirements like structural integrity, lightweight, high-temperature capability, etc. In general, the parts manufactured by FFF show large void contents, weak bonding, and inferior structural performance in comparison to those produced by conventional methods. The present research focused on overcoming the shortcomings of FFF through process modeling, microstructure study, and performance analysis. An experimental and numerical study has been conducted on the FFF of carbon fiber reinforced polylactic acid (CF/PLA) composites. A computational fluid dynamics (CFD) based numerical model was developed to simulate the temperature distribution and melt flow characteristics of highly viscous polymer (single and two-phase composites) using non-Newtonian computational model. Free space bead flow geometry and bead spreading architecture on the platform were also simulated with various nozzle geometries. The effects of the circular, square, and star-shaped geometries on bead cross-sectional shapes were studied both numerically and experimentally to achieve less void contents and improve the bead/layer bonding. Different dominant FFF process variables, both in filament extrusion and part production steps were studied, and a multi-level experimentation scheme was developed to study the bead-level to part-level properties. Physics-based surrogate models were developed, and stochastic uncertainty analysis was carried out on the manufacturing process to build up an optimum process-structure relationship. Another criticality of fiber-matrix interfacial bonding in the FFF-composites was addressed by introducing proper surface treatment to the fibers and post-manufacturing thermal treatment. The numerical model showed good promise in tailoring the bead geometry with the square and star-exit nozzle that potentially enhanced the bead to bead bonding. Extensive experimental studies were conducted to characterize strength, stiffness, fracture toughness, and void contents with various printed layer orientations and fiber concentrations of the FFF coupons. An acid-based functionalization of fibers, printing using square-nozzle, and enhanced crystallinity through controlled annealing were found to improve the fiber-matrix and inter-bead bonding, reduce the inter and intra-void and improve mechanical performances. The optimization and experimental data-driven stochastic modeling of the process parameters paved the way for producing parts with greater confidence at reduced experimental affords. The investigations and strategies developed in this dissertation will help to establish a high-quality and efficient process framework to improve the performance of additively manufactured two-phase composites. The fundamental understating and knowledge exercised in this dissertation can potentially be used for any polymer-based AM processes beyond the FFF since the fundamental challenges of controlling the voids and bonding are unavoidable.
  • Thumbnail Image
    Item
    Scalable Deep Learning-Based Quantitative Ultrasound Tomography for Medical Imaging
    (University of Alabama Libraries, 2025) Anwar, Shoaib; Su, Weihua
    Quantitative ultrasound tomography (QUT), powered by full waveform inversion (FWI), enables precise tissue differentiation and diagnosis through fast, affordable scanning without the need for sedatives or X-rays. However, the extended processing times and reconstruction errors of this physics-based method limit their application in time-sensitive areas. This dissertation aims to overcome these challenges by developing a robust cyberinfrastructure that integrates advanced artificial intelligence techniques into the FWI workflow for breast tissue characterization and lesion detection. The scarcity of clinical data is a foundational challenge in deep learning (DL) approaches. As an alternative, synthetic numerical datasets offer considerable promise. However, generating such data at a sufficient scale and fidelity is time-consuming. To address this limitation, this research develops an efficient, scalable, and high-performance computing (HPC)-based framework for the accelerated generation of FWI-based datasets of realistic numerical breast phantoms (NBPs). This framework drastically reduces dataset production timelines from several months to just a day. Building upon the datasets generated by HPC, the dissertation introduces a novel deep learning approach called adjoint theory with generative adversarial network (ATGAN). This method is designed to accelerate the FWI process. ATGAN integrates the traditional adjoint-tomography theory with a GAN architecture into FWI-based QUT by embedding strong physics-based priors into the conditional GAN. This approach significantly reduces computational demands. Comparative analyses demonstrate that ATGAN substantially outperforms classical U-Net models, demonstrating improved generalization, greater training stability, and enhanced preservation of critical structural details, particularly when reconstructing high-resolution wave speed maps from initial low-fidelity estimates. In addition to the data-driven ATGAN approach, a second method known as physics-guided neural network FWI (PNFWI) has been developed. PNFWI employs a fully unsupervised, physics-guided neural network strategy that operates directly with raw ultrasound measurement data eliminating the need for initial wave speed assumptions or ground-truth wave speed maps. Central to the PNFWI approach is the cycle-consistency loss, which enables the network to train while solving the physics of wave propagation. This significantly mitigates the risk of cycle skipping and ensures stable convergence even under conditions of added data noise. Therefore, this approach holds substantial promise for clinical applications by potentially reducing false-positive rates.
  • Thumbnail Image
    Item
    A Study of Cure Kinetics, Microstructures, and Properties of Thermoset Composites Manufactured Through Frontal Polymerization
    (University of Alabama Libraries, 2025) Shams, Aurpon Tahsin; Haque, Anwarul
    This dissertation presents a comprehensive study on the development and investigation of fast-curing epoxy composite systems utilizing frontal polymerization (FP), a self-sustaining, energy-efficient curing technique driven by the exothermic nature of polymerization reactions. The research addresses critical challenges in conventional thermoset composite manufacturing, such as prolonged curing cycles, high energy consumption, void formation, and residual stress development, by leveraging the effects of thermal (TPED-1,1,2,2-Tetraphenyl-1,2-ethanediol) and cationic initiators to accelerate the curing process without compromising material performance.A comprehensive experimental approach was used to study the cure kinetics, thermal behavior, and mechanical properties of both pure epoxy and carbon fiber reinforced epoxy composites. Adding short carbon fibers (SCFs) and applying process improvements, such as solvent-assisted dispersal of initiators and using silicone molds, led to better microstructural integrity, fewer voids, and lower thermal stresses. Differential Scanning Calorimetry (DSC) tests showed high cure levels (up to 98%), while tensile and hardness measurements indicated improved mechanical performance, with strengths reaching up to 117 ksi and notable gains in Vickers hardness.Meanwhile, an advanced computational model was created using finite element analysis to simulate the thermal and chemical behaviors of FP. It incorporates cure kinetics parameters from DSC data and employs the Chile-Alpha model for predicting residual strain. The coupled thermo-mechanical simulations accurately captured front propagation, temperature distribution, and strain development, closely matching experimental results. The effect of fiber aspect ratio, volume fraction, and orientation on curing behavior was also studied through simulation, providing valuable insights for optimizing composite design to improve curing efficiency and mechanical properties.This work combines experimental research with predictive modeling to create a robust framework for the FP-based manufacturing of thermoset composites. By focusing on key factors such as initiator concentration, heat management, fiber reinforcement, and residual strain reduction, the study promotes FP as a practical technique for additive manufacturing (AM) and rapid prototyping in aerospace, automotive, and structural applications. The results enhance the fundamental understanding of FP mechanisms while providing practical solutions for the scalable production of defect-free composites.