ENHANCED TEACHER-LEARNING BASED ALGORITHM IN REAL SIZE STRUCTURAL OPTIMIZATION

doi:10.3846/jcem.2022.16387

Return

ENHANCED TEACHER-LEARNING BASED ALGORITHM IN REAL SIZE STRUCTURAL OPTIMIZATION

2022-03-23

Mohammad Sadegh Eshaghi

Alireza Salehi

Alfred Strauß *

DOI：10.3846/jcem.2022.16387

Original

Original request for help

Save

Abstract

En 中文

Space frame structures that are made up of a huge number of members are often used on a large scale, hence their accurate evaluation is important to achieve the optimal design. On the other hand, the use of space Frames and 3D truss structures has become more popular due to its time efficiency. Also, these types of structures can carry loads in long-pan buildings and are used in large-scale structures such as halls, hangars, passenger stations, etc. In this study, a novel evolutionary algorithm, named ETLBO, has been proposed for the optimization of space frame design in real-size structures. Despite the existing methods in the literature, the ETLBO method can be used for large-scale space frame structures due to its high speed with sufficient accuracy. At first, four optimization algorithms Particle swarm optimization (PSO), Genetic Algorithm (GA), Differential Evolution (DE), and Teaching-learning-based optimization (TLBO) under structural problems have been evaluated. The results show that the TLBO algorithm performs better in solving problems and has been better in most problems than other algorithms. So, we have tried to improve this algorithm based on a machine learning approach and combination operators. Algorithm improvement is created by adding a crossover operation between the new solution and the best solution in the teacher phase. This change causes a sudden movement and escapes from the local minima for the algorithm. Enhanced algorithm results show that convergence speed and optimal response quality have improved. Finally, using this algorithm, several new practical examples have been optimized.

Keywords:

large-scale structure

space frame structures

optimization

hybrid method