Genetic Algorithm-Based Decision Support for Optimizing Seismic Response of Piping Systems

This paper describes computational approaches used in a prototype decision support system (DSS) for seismic design and performance evaluation of piping supports. The DSS is primarily based on a genetic algorithm (GA) that uses finite element analyses, and an existing framework for high performance distributed computing on workstation clusters. A detailed discussion is presented on various issues related to the development of an efficient GA implementation for evaluating the trade-off between the number of supports and cost. An integer string representation of the type used in some existing studies, for instance, is shown to be inferior to a binary string representation, which is appropriate when supports are modeled as axially rigid. A novel seeding technique, which overcomes the inefficiencies of conventional methods in the context of pipe support optimization, is also presented. Finally, an efficient crossover scheme is proposed for generating trade-off curves and the approach is validated with respect to optimal solutions obtained by enumeration. In addition to computational enhancements, the role of joint-cognitive decision making is explored using "Modeling to Generate Alternatives - MGA," a methodology based on optimization to produce alternatives that may spur creativity and offer new insights. These computational approaches are illustrated with applications to a simple, representative piping system, as well as an actual power plant piping system.