This thesis details the development and implementation of a novel design-driven harmony search (DDHS), derived from the recently introduced harmony search algorithm, in a variety of steel frame optimization problems. By using constraint evaluation data from vectors stored in the harmony matrix and using logical search neighborhoods to define the appropriate magnitude and direction of trial solution mutations, DDHS demonstrates improved performance over other stochastic algorithms as measured by the optimality of designs obtained, the frequency with which optimal designs are obtained, and the amount of required computational cost. DDHS is validated in seven standard weight minimization optimization examples. Additionally, DDHS is implemented in optimizations in which frame sections and beam fixities are variable and the objective function considers the high cost of moment connections. When paired with standard harmony search in a two-phase method proposed herein, DDHS consistently identifies optimal designs for such connection topology optimizations.