Robust Optimization with Interval Uncertainties Using Hybrid State Transition Algorithm

Robust optimization is concerned with finding an optimal solution that insensitive to uncertainties and has been widely used in solving real-world problems. However, most robust methods suffer from high computational costs poor convergence. To alleviate the above problems, improved algorithm proposed. First, reduce cost, second-order Taylor series surrogate model approximate robustness indices. Second, strengthen convergence, state transition studied explore whole search space for candidate solutions, while sequential quadratic programming adopted exploit local area. Third, balance optimality of a preference-based selection mechanism investigated which effectively determines promising solution. The proposed method applied obtain solutions seven examples are subject decision variables parameter uncertainties. Comparative studies other algorithms (robust genetic algorithm, Kriging metamodel-assisted method, etc.) show can accurate less cost.