Reliability analysis and reliability-based optimization of composite laminated plate subject to buckling

The purpose of this thesis is to investigate the effects of variations in design parameters on reliability of a composite plate subject to buckling. Then, the reliability-based design which maximize the reliability in terms of ply orientation angles of individual layers is obtained. It is illustrated the importance of considering the structural reliability in designing the composite plate subject to buckling.;The composite material is known to have more uncertainties than a conventional material due to the fabrication process. It has been known that a deterministic optimum design is strongly anisotropic and very sensitive to the change in loading conditions. Therefore, it is necessary to consider the effect of variations in design parameters by applying the structural reliability theory.;The reliability is evaluated by modeling the buckling failure as a series system consisting of potential eigenmodes. The mode reliability is obtained by the first-order reliability theory (FORM), where material constants, ply orientation angles and the applied loads are considered as random variables. In order to keep track of the intended buckling mode during the reliability analysis, the mode tracking method is utilized. Then, the failure probability of the series system is approximated by Ditlevsen's upper bound. The reliability-based optimization is formulated to find a laminate construction which maximize the system reliability. The problem is formulated as a nested problem with two levels of optimization; the reliability analysis and the reliability-based design.;Through numerical calculation, the laminate construction of the reliability-based design is shown to be much different from that of the deterministic one. The deterministic buckling maximum design have more than two critical buckling modes. However, the reliability-based design has close mode reliabilies for the critical mode. The well balanced mode reliabilities in the latter lead to a higher system reliability. Finally, effects of the correlation between the random variables are investigated. It is clarified that the reliability-based design changes significantly in terms of the correlation coefficients. It is also shown that the reliability-based design ignoring correlation is sometimes less safe than even a deterministic buckling load maximization design when the random variables are correlated.