ABSTRACT
However, the procedure described above is not very different from an automated trial-and-error procedure and, consequently, it only performs well when very large populations are employed (or, alternatively, when the search area is small). As a result, GAs employ a set of techniques designed to improve substantially the efficiency of the optimization algorithm. The first of these operations is termed “selection” and its purpose is to decide on which individuals are discarded and which are transferred to the next generation. Although many methods exist, it is generally accepted that the best should remain in the population. The second procedure, known as “crossing”, consists of determining which characteristics (i.e. genes) of the parent individuals (“ancestors”) are transmitted to the offspring (“descendants”). Finally, the third genetic operator, termed “mutation”, introduces small changes to these genes when the offspring is being determined. Once the individuals composing the new generation are
known, the previously described assessment process is restarted. Naturally, due to the evolutionary aspect of the GAs, it is expected that the best individual of a given generation corresponds to a better solution than the best individual of the previous population. In the present case, this means that the ability of a given constitutive model to reproduce a certain material behaviour increases with the number of analysed generations.
