Plate heat exchanger, approximation assisted optimization, CFD, multiobjective optimization
This paper presents a comparison between different multiobjective optimization approaches that can be used to optimize the design of thermal equipment. Plate heat exchanger is taken as case study to apply different optimization techniques. The thermal-hydrodynamic characteristics of single phase turbulent flow in chevron-type plate heat exchangers with sinusoidal-shaped corrugations have been used in this paper. The computational domain contains a corrugation channel and the simulations adopted the shear-stress transport (SST) κ-ω model as the turbulence model. Two different approximation assisted optimization approaches are tested. Offline approximation assisted optimization, and online approximation assisted optimization are compared to optimize plate heat exchanger design. For both approximation techniques (offline and online), design optimization is performed using multiobjective genetic algorithm based on meta-models that are built to represent the entire design space. In offline approximation, globally accurate meta-models are built which requires adding more samples. However in online approximation assisted optimization, samples are added just to improve the metamodels performance in the expected optimum region. Approximated optimum designs are validated using computationally expensive actual CFD simulations. Finally, a comparison between offline and online approximation assisted optimization is presented with guidelines to apply both approaches in the area of heat exchanger design optimization. The methods presented in this paper are generic and can be applied to optimize different types of heat exchangers, electronic cooling devices and other thermal system components.