Improved Teaching-Learning-Based Optimization Algorithm And Its Application To Optimal Design Of Trusses And Steel Frames

With the development of technology and equipment, the optimal design model of engineering structure has gradually shifted from a simplified model based on a number of assumptions to a complex model which reflects the real structure; this change improves the accuracy and applicability of the optimization. But the optimization based on a complex model features high-dimensional variables, multi-equations, nonlinear, etc. This makes it difficult to solve the optimization and store variables. In the face of such optimization problems, traditional optimization methods have deficiencies in initial sensitivity and convergence, and then we need an efficient optimization method to make up for the deficiencies of traditional optimization methods. In recent years, bio-inspired computing inspired by the biological behavior has many advantages, such as easy to implement, efficient optimize performance, calculation is relatively simple and no problem specific information, which make bio-inspired computing play an important role in solving complex engineering structural optimization.Firstly, this paper reviews the development of bio-inspired computing and its application to the structural optimal design. Secondly, this paper introduces a new bio-inspired algorithms—Teaching-Learning-Based Optimization (TLBO), then apply it in cross-sectional optimization of space trusses and geometry optimization of trusses. Some problems are found through its performance in the structural optimization, that is, setting elite strategy makes no significant difference among improving performance of optimization algorithms, moreover, TLBO is easy to fall into local optimum when variables are low-dimensional and solution space is complex. In order to make up for these deficiencies, this paper introduces the chaos search in TLBO, with the ergodicity and stochastic property of chaos, the algorithm can jump out of the local optima, increasing the ability of getting the optimal solution. Besides, the teacher phase of TLBO only uses the population average, which results in lack of interactivity between the populations. This strategy is easy to cause the algorithm premature. Information can be transferred among individuals of the swarm by introducing passive congregation to TLBO. After a series of improvements, the improved algorithm is called Modified chaotic Teaching-Learning-Based Optimization (MCTLBO).The improved algorithm inherits the advantages of the old algorithm that does not need specific parameters. Finally, MCTLBO is tested with the discrete variables cross-sectional optimization of steel frames and the mixed variables dynamic optimal design of trusses. Experimental results indicate that the MCTLBO has a better search performance than TLBO. It is much desired for MCTLBO to be applied to the tasks of optimal design of engineering structures.