| Popis: |
In recent years most automotive companies have become much more serious about using carbon fiber composites for light-weighting automotive structures. In order to efficiently design the structures to perform at or above current safety standards, engineers look towards predictive numerical simulation using experimentally determined material constants. However, many challenges hinder the wide use of composite materials in design of structural components. This thesis focuses on the following three challenges: the wide range of possible choices for composite systems; the complexity of the damage initiation and propagation; and the numerous experiments required for full characterization of each new material system.This work explores the use of microscopic images of the material’s cross section for measuring the fiber volume fraction. This technique should allow for better evaluation of the different micromechanics formulae. Obtaining more reliable micromechanics formulae should mitigate the first design challenge, as it allows approximating the mechanical properties and facilitates comparing many composite systems with relatively limited required information. The fiber volume fractions of several samples were measured using the presented technique, and the results are evaluated.To address the complexity of the damage initiation and propagation, unsupervised learning techniques were applied to the AE data. The resulting AE groups were then related to the different damage mechanisms based on experimental observations. This study aims at understanding the complex damage mechanisms and is a step towards the potential real-time damage monitoring for the assessment of the damage state. Strong correlations between the behaviors of the AE groups and the damage mechanisms are observed and discussed.Inverse methods are explored as a solution to overcome the numerous experiments required for full characterization. The advancement of full-field deformation measurement techniques allows better performance of inverse methods for determination of elastic constants. Two numerical indices for evaluating the performance of inverse methods with different sample geometries is proposed and applied to several candidate shapes. Two versions of Finite Element Model Update (FEMU) approaches were investigated and their performance compared. The methods were also applied to experimental data from testing unidirectional carbon fiber reinforced epoxy, and results were compared to values obtained from standard testing. Also, decomposition of the stiffness matrix is proposed for simplifying the forward problem reducing the required calculation time. |
