Abstract
Heat exchanger performance is strongly influenced by the refrigerant circuitry, i.e., the connection sequence of the tubes. This article describes an evolutionary computation-based approach for designing an optimized refrigerant circuitry used in an intelligent system for heat exchanger design. The technique used in this design employs two methods to generate designs implemented separately in two modules: the knowledge-based evolutionary computation module and the symbolic-learning-based evolutionary computation module. The optimization example presented in this article employed each module independently and used the combined approach to demonstrate the performance of each module and the power of the combined module approach. The best circuitry designs determined through these optimization runs yielded substantial capacity improvements over the original design; the symbolic-learning- and knowledge-based modules returned circuitry designs that improved the heat exchanger capacity by 2.6% and 4.8%, respectively, while the combined module approach resulted in a circuitry design that improved the capacity by 6.5%.
Acknowledgments
This article not subject to U.S. copyright law. David A. Yashar, PhD, Associate Member ASHRAE, is Mechanical Engineer. Janusz Wojtusiak, PhD, is Assistant Professor and Director. Kenneth Kaufman, PhD, is Computer Scientist. Piotr A. Domanski, PhD, Fellow ASHRAE, is Leader, HVAC&R Equipment.