Cost optimization of plate girders by dynamic programming
Heithecker, Gary Lee
Holt, E. C.
Master of Science
An algorithm is developed for the optimum design of a single span steel plate girder with equal flanges and unstiffened web. The height of the girder varies linearly. The flange width, flange thickness, and web thickness are discrete variables. The algorithm is formulated as a multilevel decision making process. The optimum values of the design variables are obtained subject to nonlinear behavior functions limiting stresses and deflection, and side constraints restricting the range and independence of the variables. A nonlinear objective function involving cost of material and fabrication is used as the criterion for optimization. Optimization procedures are formulated using dynamic programming. The characteristics of the algorithm are demonstrated by a number of design examples, and the designs produced by the algorithm are related to the traditional design procedure.