OPTIMAL STEEL FRAME DESIGN BY SIMULATED ANNEALING

A simulated annealing strategy is developed for use in the discrete optimization of three-dimensional steel frames. This strategy randomly perturbs the current design to create a candidate design. A probabillistic acceptance criterion is then employed to determine whether the candidate design should replace the current design or be rejected. This acceptance criterion allows worse designs to be accepted in the initial stages of the strategy. The likelihood of accepting worse designs is small in the final stages of the strategy. The strategy is presented and illustrated on a three-dimensional, six-story, unsymmetrical frame. The frame is realistically loaded with gravity and seismic loads. Members in the frame must be selected from among discrete standarized shapes. The strategy is able to treat multiple section properties per member without having to curve-fit dependent properties as functions of a single independent property. Performance of the strategy is compared to that of the branch-and-bound method. Approximation techniques aimed at reducing computation time are investigated.