Parameter Design Strategies: A Comparison Between Human Designers and the Simulated Annealing Algorithm

Abstract

Computer-based tools have great potential for facilitating the design of large-scale engineering systems. Interviews with veteran designers of desalination systems revealed that they tended to employ a trial-and-error approach to determine critical design parameters when using software design packages. A series of human experiments were conducted to observe the performance and behavior of test subjects during a series of simulated design processes involving seawater reverse osmosis (SWRO) plants. The subjects were mostly students with a spectrum of knowledge levels in desalination system design. The experiments showed that subjects who ranked top in performance behaved very differently from those who were bottom-ranked. The problem-solving profiles of the best performing subjects resembled a well-tuned simulated annealing optimization algorithm while the worst performing subjects used a pseudo random search strategy. This finding could be used to improve computerbased design tools by utilizing the synergy between strengths of humans and computers.