LAUSR

Abstract In this paper, surrogate-assisted optimisation of a U-shaped baffle square-duct heat exchanger (USBSD-HX) is proposed. A single-objective design problem is formulated to optimise the heat exchanger thermal performance and friction loss. Several surrogate models including a Kriging model, radial basis function interpolation, a quadratic response surface model and support vector regression are used for objective function approximation. The self-adaptive differential evolution with neighbourhood search (SaNSDE) is improved by introducing a new mutation strategy and employing the weighted sum of exploitation and exploration indicators for solution selection. A design demonstration of USBSD-HX reveals that the simultaneous use of multiple surrogate models gives better results. The proposed meta-heuristic is a powerful optimiser for this design task. Optimum USBSD-HX using air as the working fluid is also obtained in this study.