UBC Theses and Dissertations
Reliability-based optimization of plywood-web beams Menun, Charles Alexander
The optimal design of a plywood-web beam, as for any structural element, is usually found by trial and error in which an initial design is modified until a solution which maximizes the beam's efficiency and meets a set of prescribed design criteria is found. To automate this process, a reliability-based optimization program which computes the optimal dimensions of a plywood-web beam is formulated and tested in this study. The program minimizes the cost of a plywood-web beam subject to constraints imposed upon its performance expressed in terms of acceptable levels of safety with respect to a set of limit states. A plywood-web beam model which incorporates the effects of shear deformations in the web components and the effects of non-rigid connections between the beam's flanges and webs is developed and used to compute a plywood-web beam's response under load. The beam's performance is evaluated by means of a reliability analysis in order to rationally account for any uncertainty associated with the beam's material properties and the loads acting on it. An existing non-linear optimization routine computes the optimal continuous design of a plywood-web beam using the results of the structural and reliability analyses. A discrete solution, representing the optimal practical design which uses only available or allowable dimensions for the beam's components, is found by means of an exhaustive search of a restricted region of the design variable space near the optimal continuous solution. As an example, the program is used to optimize the design of a ply wood-web box girder. Sensitivity analyses are performed on the optimal design in order to identify and quantify the critical input parameters in the optimization process. The effects of errors in the problem's formulation, analytical errors arising from the structural analysis and statistical errors resulting in an inaccurate representation of the problem's probabilistic characteristics are studied.
Item Citations and Data