Last modified: 2021-05-20
Abstract
Searching more diverse and uniform optimization solutions has always been one of the important research directions in multi-objective optimization. To alleviate the expensive computational cost and ensure the solution diversity and uniformity, a high IGD (Inverted generational distance) index driven surrogate-based multi-objective optimization method is proposed, which can be able to update the surrogate model sequentially and guarantee a competitive optimization solutions by infilling new samples intelligently.
In the proposed method, a low-fidelity surrogate model is built for the objective functions using the relatively small number of initial samples. Then, a most promising design point is generated into the sampling pool with intelligent exploration of both the objective space and design space, employing HVEI-EGO (Hyper-volume expected improvement-Efficient global optimization, HVEI-EGO) and Euclidean distance selection strategy, respectively. To pursue higher IGD, the addition of new samples are terminated until the Pareto front points reaches pre-set numbers of Pareto front points and the optimization reaches the maximum number of expensive function simulations. Several numerical examples are investigated to demonstrate the effectiveness of the proposed method, the results show the effectiveness of additive samples.