Search results

Although genetic algorithm (GA) has already been extended to various types of engineering problems, tuning its parameters is still an interesting field of interest. Some recent works have addressed attempts requiring several GA runs, while more interesting approaches aim to obtain proper estimate of a tuned parameter during any run of genetic search. This paper seeks to address this issue.

In this paper, a competitive frequency‐based methodology is proposed to explore the least proper population size as a major affecting control parameter of GAs. In the tuning stage, the indirect shared memory in ant strategies is borrowed in a discrete manner to generate a dynamic colony of the most successive recent solutions to be added into each new population. An adaptive variable band mutation based on direct index coding for structural problems is also employed to increase the convergence rate as well as to prevent premature convergence especially after determining a proper population size. As an important field of engineering problems, the method is then applied to a number of structural size and layout optimization examples in order to illustrate and validate its capability in capturing the problem optimum with reduced computational effort.

It was shown that improper fixed size population can lead to premature convergence. Applying the proposed method could result in a more efficient convergence to the global optimum compared with the fixed size population methods.

A novel combination of genetic and ant colony approaches is proposed to provide a dynamic short‐term memory of the sampled representatives which can enrich the current population, avoiding unnecessary increase in its size and the corresponding computational effort in the genetic search. In addition, a dynamic band mutation is introduced and matched with such a search, to make it more efficient for structural purposes.

The generality of the genetic search in the light of proper coding schemes, together with its non‐gradient‐based search, has made it popular for many discrete problems including structural optimization. However, the required computational effort increases as the cardinality of the search space and the number of design variables increase. Memetic algorithms are formal attempts to reduce such a drawback for real‐world problems incorporating some kind of problem‐specific information. This paper aims to address this issue.

In this paper both Lamarckian and Baldwinian approaches for meme evolution are implemented using the power of graph theory in topology assessment. For this purpose, the concept of load path connectivity in frame bracing layouts is introduced and utilized by the proposed graph theoretical algorithms. As an additional search refinement tool, a dynamic mutation band control is recommended. In each case, the results are studied via a set of ultimate design family rather than one pseudo optimum. The method is further tested using a number of steel frame examples and its efficiency is compared with conventional genetic search.

Here, the problem of bracing layout optimization in steel frames is studied utilizing a number of topological guidelines.

The method of this paper attempts to reduce the computational effort for optimal design of real‐world problems incorporating some kind of problem‐specific information.

Genetic Algorithm, as a generalized constructive search method, has already been applied to various fields of optimization problems using different encoding schemes. In conventional GAs, the optimum solution is usually announced as the fittest feasible individual achieved in a limited number of generations. In this paper, such a pseudo‐optimum is extended to a neighborhood structure, known as optimal design family.

In this paper, the constructive feature of genetic search is combined with trail update strategy of ant colony approach in a discrete manner, in order to sample more competitive individuals from various subspaces of the search space as a dynamic‐memory of updating design family.

The proposed method is applied to structural layout and size optimization utilizing an efficient integer index encoding and its appropriate genetic operators. Different applications of the proposed method are illustrated using three truss and frame examples. In the first example, topological classes are identified during layout optimization. In the second example, an objective function containing the stress response, displacement response, and the weight of the structure is considered to solve the optimal design of non‐braced frames. This approach allows the selection of less sensitive designs among the family of solutions. The third example is selected for eigenvalue maximization with minimal number of bracings and structural weight for braced frames.

In this paper, a pseudo‐optimum is extended to a neighborhood structure, known as optimal design family.

Layout optimization of structures aims to find the optimal topology and member sizes in an integrated manner. For this purpose, the most successful attempts have addressed the outstanding features of the genetic algorithms.

This paper utilizes a direct index coding (DIC) in a way that the optimization algorithm can simultaneously integrate topology and size in a minimal length chromosome in order to seek the true optimum in an efficient and reasonable manner. Proper genetic operators are adopted for this special kind of encoding together with some modifications in the topological mutation aiming to improve the convergence of the algorithm.

The present DIC, has the following features: enforcing one‐to‐one correspondence between discrete genotype space and the problems' phenotype space; avoiding any out‐of‐bound parameter addressing and limiting the GA search only to necessary genotypes; reduction in the size of genotype search space to increase the algorithm convergence and the possibility of leading to the global optimum; dealing with direct genetic operators so that the GA parameters can be purely controlled to tune the desired balance between convergence and escaping from local optima.

Employing direct index chromosome makes it possible to eliminate the additional topological bits in treated examples.

The purpose of this paper is to present a modified version of the non-dominated sorted genetic algorithm with an application in the design optimization of a power inductor for magneto-fluid hyperthermia (MFH).

The proposed evolutionary algorithm is a modified version of migration-non-dominated sorting genetic algorithms (M-NSGA) that now includes the self-adaption of migration events- non-dominated sorting genetic algorithms (SA-M-NSGA). Moreover, a criterion based on the evolution of the approximated Pareto front has been activated for the automatic stop of the computation. Numerical experiments have been based on both an analytical benchmark and a real-life case study; the latter, which deals with the design of a class of power inductors for tests of MFH, is characterized by finite element analysis of the magnetic field.

The SA-M-NSGA substantially varies the genetic heritage of the population during the optimization process and allows for a faster convergence.

The proposed SA-M-NSGA is able to find a wider Pareto front with a computational effort comparable to a standard NSGA-II implementation.

The computational drawbacks of existing numerical methods have forced researchers to rely on heuristic algorithms. Heuristic methods are powerful in obtaining the solution of optimization problems. Although they are approximate methods (i.e. their solution are good, but not provably optimal), they do not require the derivatives of the objective function and constraints. Also, they use probabilistic transition rules instead of deterministic rules. The purpose of this paper is to present an improved ant colony optimization (IACO) for constrained engineering design problems.

IACO has the capacity to handle continuous and discrete problems by using sub‐optimization mechanism (SOM). SOM is based on the principles of finite element method working as a search‐space updating technique. Also, SOM can reduce the size of pheromone matrices, decision vectors and the number of evaluations. Though IACO decreases pheromone updating operations as well as optimization time, the probability of finding an optimum solution is not reduced.

Utilizing SOM in the ACO algorithm causes a decrease in the size of the pheromone vectors, size of the decision vector, size of the search space, the number of function evaluations, and finally the required optimization time. SOM performs as a search‐space‐updating rule, and it can exchange discrete‐continuous search domain to each other.

The suitability of using ACO for constrained engineering design problems is presented, and applied to optimal design of different engineering problems.

The purpose of this paper is to develop new constitutive models to predict the soil deformation moduli using multi expression programming (MEP). The soil deformation parameters formulated are secant (Es) and reloading (Er) moduli.

MEP is a new branch of classical genetic programming. The models obtained using this method are developed upon a series of plate load tests conducted on different soil types. The best models are selected after developing and controlling several models with different combinations of the influencing parameters. The validation of the models is verified using several statistical criteria. For more verification, sensitivity and parametric analyses are carried out.

The results indicate that the proposed models give precise estimations of the soil deformation moduli. The Es prediction model provides considerably better results than the model developed for Er. The Es formulation outperforms several empirical models found in the literature. The validation phases confirm the efficiency of the models for their general application to the soil moduli estimation. In general, the derived models are suitable for fine‐grained soils.

These equations may be used by designers to check the general validity of the laboratory and field test results or to control the solutions developed by more in‐depth deterministic analyses.

The purpose of this paper is to propose a novel improved teaching and learning-based algorithm (TLBO) to enhance its convergence ability and solution accuracy, making it more suitable for solving large-scale optimization issues.

Utilizing multiple cooperation mechanisms in teaching and learning processes, an improved TBLO named CTLBO (collectivism teaching-learning-based optimization) is developed. This algorithm introduces a new preparation phase before the teaching and learning phases and applies multiple teacher–learner cooperation strategies in teaching and learning processes. Applying modularization idea, based on the configuration structure of operators of CTLBO, six variants of CTLBO are constructed. For identifying the best configuration, 30 general benchmark functions are tested. Then, three experiments using CEC2020 (2020 IEEE Conference on Evolutionary Computation)-constrained optimization problems are conducted to compare CTLBO with other algorithms. At last, a large-scale industrial engineering problem is taken as the application case.

Experiment with 30 general unconstrained benchmark functions indicates that CTLBO-c is the best configuration of all variants of CTLBO. Three experiments using CEC2020-constrained optimization problems show that CTLBO is one powerful algorithm for solving large-scale constrained optimization problems. The application case of industrial engineering problem shows that CTLBO and its variant CTLBO-c can effectively solve the large-scale real problem, while the accuracies of TLBO and other meta-heuristic algorithm are far lower than CLTBO and CTLBO-c, revealing that CTLBO and its variants can far outperform other algorithms. CTLBO is an excellent algorithm for solving large-scale complex optimization issues.

The innovation of this paper lies in the improvement strategies in changing the original TLBO with two-phase teaching–learning mechanism to a new algorithm CTLBO with three-phase multiple cooperation teaching–learning mechanism, self-learning mechanism in teaching and group teaching mechanism. CTLBO has important application value in solving large-scale optimization problems.

The purpose of this paper is to propose an optimum design of a corona ring for insulator strings using optimisation techniques, which are gravitational search algorithm (GSA) and imperialist competitive algorithm (ICA).

An insulator string model geometry with a corona ring was modelled in a finite element analysis software, and it was used to obtain the electric field distribution in the model. The design was optimised using GSA and ICA. The variables were the corona ring diameter, ring tube diameter and vertical position of the ring along the insulator string.

Using optimisation method, the minimum electric field magnitude on the insulator string with a corona ring design is lower than without using optimisation method. GSA yields better results than ICA in terms of the optimised corona ring design.

The proposed methods can help in improvement of corona ring design in reducing the electric field magnitude on the energised end of insulator strings.

A new method to design an optimum corona ring for insulator strings, which is using optimisation methods, has been developed in this work.

Real-world applications in engineering and other fields usually involve simultaneous optimization of multiple objectives, which are generally non-commensurable and conflicting with each other. This paper aims to treat the transformer design optimization (TDO) as a multiobjective problem (MOP), to minimize the manufacturing cost and the total owing cost, taking into consideration design constraints.

To deal with this optimization problem, a new method is proposed that combines the unrestricted population-size evolutionary multiobjective optimization algorithm (UPS-EMOA) with differential evolution, also applying lognormal distribution for tuning the scale factor and the beta distribution to adjust the crossover rate (UPS-DELFBC). The proposed UPS-DELFBC is useful to maintain the adequate diversity in the population and avoid the premature convergence during the generational cycle. Numerical results using UPS-DELFBC applied to the transform design optimization of 160, 400 and 630 kVA are promising in terms of spacing and convergence criteria.

Numerical results using UPS-DELFBC applied to the transform design optimization of 160, 400 and 630 kVA are promising in terms of spacing and convergence criteria.

This paper develops a promising UPS-DELFBC approach to solve MOPs. The TDO problems for three different transformer specifications, with 160, 400 and 630 kVA, have been addressed in this paper. Optimization results show the potential and efficiency of the UPS-DELFBC to solve multiobjective TDO and to produce multiple Pareto solutions.