Research on a surrogate model updating-based efficient multi-objective optimization framework for supertall buildings

In order to solve the multi-objective optimization problems of supertall buildings (such as structural design optimization, aerodynamic shape optimization, etc.) with sizable design space more effectively, it is necessary to develop an efficient multi-objective optimization method. Therefore, generalized regression neural network optimized by genetic algorithm (GA-GRNN) based surrogate model was constructed, and a multi-objective optimization framework based on the non-dominated sorting genetic algorithm (NSGA-II) and GA-GRNN surrogate model updating was proposed. The feasibility of multi-objective optimization framework based on surrogate model updating was verified by using the experimental wind pressure data of a supertall building model, and the influencing factors of optimization efficiency were analyzed. The results show that the proposed framework has satisfactory optimization accuracy and efficiency. The optimal sample data set proportional distribution (training set: verification set: test set, i.e., T: V: T) is 7:2:1. With the increase of the total number of sample points in the design space, the optimal proportion of the initial sample points decreases significantly. A thorough consideration of the acquisition time of a single sample value and the optimal proportion of initial sample points is helpful to improve the multi-objective optimization efficiency further. Therefore, for the optimization problems in engineering applications (especially supertall buildings), it is suggested that the reasonable proportion of initial sample points of the surrogate model should be determined according to the acquisition time of a single sample value and the total number of sample points in the design space. The framework is more suitable for complex problems with large total number of sample points in design space and long acquisition time of a single sample value. This study can provide a valuable reference for further research or efficient solution to multi-objective optimization problems in practical engineering applications (such as the optimization problem of supertall buildings).

Automated summary

In order to solve the multi-objective optimization problems of supertall buildings, an efficient multi-objective optimization method was developed using a surrogate model based on genetic algorithm optimized generalized regression neural network. The proposed framework, based on non-dominated sorting genetic algorithm and GA-GRNN surrogate model updating, was verified using experimental wind pressure data and analyzed for factors influencing optimization efficiency. The framework showed satisfactory optimization accuracy and efficiency, with the optimal proportion of initial sample points determined by the acquisition time of a single sample value and the total number of sample points.