China
Africa
Explore chapters and articles related to this topic
Evolutionary Optimization Techniques as Effective Tools for Process Modelling in Food Processing
Published in Surajbhan Sevda, Anoop Singh, Mathematical and Statistical Applications in Food Engineering, 2020
Lakshmishri Roy, Debabrata Bera, Vijay Kumar Garlapati
An evolutionary algorithm is a biologically inspired, generic, population-based optimization algorithm. Its mechanism includes: Reproduction/procreation: The process of producing new “offspring” from their “parents”.Mutation: Alteration in the order of the process being considered (e.g., organism, production or business process, code).Recombination: A process of exchange of information between two processes yielding a new combination of processes (e.g., operations in a workflow process).Selection: A method by which traits become either more or less common in a population as a function of the influence of traits concerning the intended goal (e.g., increased production efficiency in a production process). Selection is a key evolution mechanism. Probable solutions of the optimization problem for which an evolutionary algorithm is employed to arrive at, are viewed as entities in a population. A fitness function is used to assess its suitability as a solution. A fitness function is an objective function that is used to summarize how close a given solution is to fulfilling the optimization goals. All the stated operators are applied several times in the process and, hence, the term “evolutionary”.
Models and Algorithms for Machine Scheduling with Setup Times
Published in Cornelius Leondes, Computer-Aided Design, Engineering, and Manufacturing, 2019
Evolutionary algorithms, or genetic algorithms, are search heuristics whose operation is modeled after natural evolutionary processes. A particular solution, or individual, is represented by a string, each element in a string determining some feature of the solution. Hence, the function of this string is analogous to that of genetic material in living organisms, and the string can be referred to as a chromosome comprised of a series of genes, each of which may take one of a range of allowed values (or alleles). We can also refer to this string as the genotype of the individual. A gene might correspond to a sequence position in a genetic algorithm for a machine scheduling problem, and the value of this gene will indicate the job to be sequenced at this position. Alternative encodings of schedules are commonplace, however. The schedule that results from interpretation of an individual's chromosome is termed the phenotype of this individual.
Introduction
Published in Joseph Y.-T. Leung, Handbook of SCHEDULING, 2004
The evolutionary algorithm developed is based on what Kling and Banerjee called simulated evolution for the placement problem in electronic circuit design [57]. Evolutionary algorithms usually evolve a population of solutions. Simulated evolution is different; it evolves a population of solution components. In driver scheduling, a driver schedule is a solution and driver shifts are its components. The algorithm iteratively operates on a single schedule starting with an initial schedule. In each iteration, individuals in the population of shifts are ranked using fuzzy evaluation discussed above. Those shifts below a predefined threshold level of fitness will be marked for deletion from the schedule. Some markings may be randomly reversed as a kind of mutation. After a portion of the schedule is duly deleted, a greedy selection heuristic similar to that in GACT discussed in Section 51.3.2.2.1 is used to repair the schedule using the large pool of candidate shifts. Fuzzy evaluation is also applied on the candidate shifts in the greedy selection heuristic. Hence, each iteration mimics a generation of evolution, the process of which is simulated in the sense that the state of the population is transformed in one go rather than through a series of repeated genetic operations such as crossover and mutation.
Multiobjective optimisation of absorption heat pump performance with double internal irreversibility based algorithm NSGAII: case of three-heat-source model
Published in International Journal of Ambient Energy, 2022
Paiguy Armand Ngouateu Wouagfack, Germaine Mabou Ninkam, Réné Tchinda
Evolutionary algorithms are algorithms based on the theory of evolution and natural selection. Their resolution has been a challenge for researchers for a long time (Coello, Lamont, and Van-Veldhuizen 2007; Deb and Jain 2013). The fundamental principle of evolutionary algorithms is to adapt individuals to their place or environment to survive and reproduce by transmitting their genes to descendants. Several scientific researchers were inspired by the term Evolutionary Computation, especially John Holland in 1975 in the study of the genetic algorithm. Figure 3 illustrates the principle of evolutionary algorithms based essentially on selection, crossover and mutation operators. This algorithm is the most popular and most widely used evolutionary algorithms. It is inspired by the biological evolution of living species, implementing the natural selection and reproduction phenomena stemming from Charles Darwin's theory. All individuals form the parents’ population, which will evolve during iterations, called ‘generations’, until a stopping criterion is verified.
An elitist cooperative evolutionary bi-level multi-objective decomposition-based algorithm for sustainable supply chain
Published in International Journal of Production Research, 2022
Malek Abbassi, Abir Chaabani, Nabil Absi, Lamjed Ben Said
Evolutionary algorithms are population-based meta-heuristics that aim at evolving the population of solutions from one iteration to another. In this context, we highlight the necessity of starting with a promising initial population. Hence, a number of researchers choose to randomly generate the initial population set. While, others consider a good heuristic or a good policy to start with high-quality solutions. Other attempts focused on starting the search with diversified solutions over the decision search space. This diversification procedure was firstly used by Das and Dennis (1998) in the context of continuous many-objective optimisation problems. Chaabani et al. (2015) proposed a new variant of the Das and Dennis (1998) method able to generate a set of solutions over the whole discrete decision search space. In our method, we aim to ensure a uniform coverage of the solution space by generating well-diversified solutions. In this way, we propose a discrete dissimilarity decomposition-based heuristic detailed in Algorithm 3 as follows:
Application of bi-objective genetic programming for optimizing irrigation rules using two reservoir performance criteria
Published in International Journal of River Basin Management, 2021
Parisa-Sadat Ashofteh, Omid Bozorg-Haddad, Hugo A. Loáiciga
Evolutionary algorithms select the best members of one population of solutions to transfer them to the next generation. This selection is not simple given that there are several objective functions in the multi-objective problems. Therefore, the selection of the best solutions is based on the notion of ranking, whereby the solutions are evaluated and ranked based on non-domination. The ranking of the solutions proceeds by comparing the objective function values of each two members of a population. The solutions that are found to be non-dominated receive a rank of 1 (F1) and are placed on the first front. Subsequently, regardless of the impact on members located in the first front, a series of other non-dominate solutions is determined and receive a rank of 2 (F2) and placed on the second front. Thus, all solutions are ranked based non-domination. This process is repeated until all the current solutions are ranked and placed on Pareto fronts.