A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Research on Comparative of Multi-Surrogate Models to Optimize Complex Truss Structures
Surrogate models have been proven to be reliable and effective methods for the application of engineering problems. This article presents a comparative study of three common surrogate models Polynomial Response Surface (PRS), Radial Basis Function (RBF) neural network, and Kriging model in terms of optimization. For the optimization of plane 10-bar truss structure and 25 bar space truss, the effectiveness of the surrogate model algorithm is verified by different surrogate models. Finally for a typical complex truss structure such as lattice boom of crawler crane, the Optimal Latin Hypercube Design (OLHD) is used to sample the optimized sample points for fitting and interpolation of three surrogate models and analyzing their errors, in order to get better surrogate effect, PRS is combined with RBF Neural Network, and secondly, the global optimization algorithm (Multi-Island Genetic Algorithm, MIGA) and gradient algorithm (Modified Method of Feasible Directions, MMFD) are used to optimize the fitted four surrogate models. Through the comparison of the optimization results, the optimization of PRS-RBF combined surrogate model using MIGA-MMFD algorithm instead of finite element model optimization has good stability and reliability. The total mass of the optimized model has been reduced by 24.47%. The number of optimization iterations is within 250 generations. The new method proposed in this paper can greatly promote the reduction of the period of analysis and optimization of engineering structures.
Research on Comparative of Multi-Surrogate Models to Optimize Complex Truss Structures
Surrogate models have been proven to be reliable and effective methods for the application of engineering problems. This article presents a comparative study of three common surrogate models Polynomial Response Surface (PRS), Radial Basis Function (RBF) neural network, and Kriging model in terms of optimization. For the optimization of plane 10-bar truss structure and 25 bar space truss, the effectiveness of the surrogate model algorithm is verified by different surrogate models. Finally for a typical complex truss structure such as lattice boom of crawler crane, the Optimal Latin Hypercube Design (OLHD) is used to sample the optimized sample points for fitting and interpolation of three surrogate models and analyzing their errors, in order to get better surrogate effect, PRS is combined with RBF Neural Network, and secondly, the global optimization algorithm (Multi-Island Genetic Algorithm, MIGA) and gradient algorithm (Modified Method of Feasible Directions, MMFD) are used to optimize the fitted four surrogate models. Through the comparison of the optimization results, the optimization of PRS-RBF combined surrogate model using MIGA-MMFD algorithm instead of finite element model optimization has good stability and reliability. The total mass of the optimized model has been reduced by 24.47%. The number of optimization iterations is within 250 generations. The new method proposed in this paper can greatly promote the reduction of the period of analysis and optimization of engineering structures.
Research on Comparative of Multi-Surrogate Models to Optimize Complex Truss Structures
KSCE J Civ Eng
Yang, Chongjian (author) / Yang, Junle (author) / Qin, Yixiao (author)
KSCE Journal of Civil Engineering ; 28 ; 2268-2278
2024-06-01
11 pages
Article (Journal)
Electronic Resource
English
Research on Comparative of Multi-Surrogate Models to Optimize Complex Truss Structures
| Springer Verlag | 2024
Surrogate measures to optimize structures for robust and predictable progressive failure
| British Library Online Contents | 2009
| British Library Online Contents | 2011
Planar Truss Structures with Multi-Symmetry
| British Library Conference Proceedings | 2008