Computational Approaches for Decision Support in Structural Performance Evaluation

The objective of this paper is to promote the development of new hybrid optimization methods that combine simple algorithms based on structural engineering knowledge with sophisticated optimization algorithms. We illustrate the importance of this concept using algorithms that we have developed for different problems in structural engineering. First, we present a new algorithm for truss optimization that demonstrates a significant improvement compared to existing methods such as GAs. The performance of the algorithm is studied using widely popular examples in literature like the 10-bar truss. Then, we describe a GA-based approach with special adaptation for performing trade-off studies on discrete optimization problems with binary decision variables. The approach employs a unique crossover scheme that ensures that the offspring from the crossover have the same number of ones given that the two parents each had the same number of ones. The superior performance of this crossover over the traditional uniform crossover for trade-off studies is illustrated with application to certain examples. Finally, "MGA - Modeling to Generate Alternatives," (MGA) is used to generate alternatives that are closely-spaced in objective space but are farther-apart in decision space. MGA is different from GA in the sense that GAs evaluate alternatives only in objective space but not in decision space.