20 September 2011
In this paper an online multiobjective approximation assisted optimization
approach is used to design a novel air-cooled heat exchanger using multiscale simulation. Design optimization is performed using multiobjective genetic algorithm while the computational cost was reduced significantly by applying an online approximation technique.
Higher model fidelity is achieved by applying the multiscale heat exchanger simulation method. The approach is based on the principle that when more points are sampled in the vicinity of the expected Pareto frontier (optimum design region), a fewer number of sample points will be eventually required for building a reasonably accurate metamodel. This approach uses a CFD technique coupled with e-NTU solver for heat exchanger performance evaluation. Comparing the results with offline approximation assisted optimization, the proposed online approximation assisted optimization approach resulted in better optimum results while reducing the computational time significantly.