Description
Composite materials offer numerous advantages over traditional monolithic materials such as higher specific strength, greater specific stiffness, greater corrosion resistance and good vibration damping characteristics. Improved use of composite materials is a key element to improving energy efficiency of future plane, trains and automobiles. One of the challenges to overcome for achieving higher efficiencies with composite materials is the development of more accurate models and material parameters for composite failure prediction. There are currently over twenty failure criteria to predict failure of fiber reinforced polymer matrix composite materials. Recent assessment of failure prediction models in a worldwide failure exercise showed that the model formulated by Puck (1994) to be the most accurate. However, the Tsai-Wu criterion that was developed in the 1960 and 70s continues to be the work horse in industrial applications. Existing models fail to predict composite material failure accurately under certain loading regimes. To overcome this deficiency engineers simply choose larger safety factors that leads to heavier designs. Using a more accurate model, can allow us to decrease these safety factors and thereby achieve lighter structural designs. The recently formulated Puck's failure model provides such accuracy. However, the model requires as many as eleven material parameters. Some of these parameters require multiplicity of tests/experiments to deduce them. Obtaining material parameters from tests and quantifying their variability is a costly and time consuming exercise. Inaccurately quantified material parameters require enforcing larger safety factors in the design to safeguard for the material property data uncertainty. Gains in improving model fidelity are negated by the inaccuracy of the material parameters available. This thesis presents the results of a global variance sensitivity of material parameters needed for composite failure prediction using Puck's failure theory. Global variance sensitivities quantify the relative contribution of variances in individual material parameters to the total variance of the laminate failure load. The use of total variance sensitivity to identify optimal laminates and loading conditions needed to characterize new material properties more accurately is demonstrated.
