Keywords: genetic algorithm, identification, mechanical system.

Finding a model and its parameters starting from a collection of measured data is a very important challenge for engineers and researchers. Systems identification is one of the most classical problems in mechanical sciences because it aims at developing a bridge between theoretical models (being of a mathematical nature and limited by the our knowledge of the phenomena) and the real world. By virtue of its crucial role in engineering and applied physics, the scientific community has spent great effort in the last decades to improve the reliability and applicability of the available theories. As a consequence, the interested reader can find abundant literature concerning this topic.

Nonetheless, there is a wide number of fields in which this problem remains one of the most complicated to solve. The existence of the solution and its uniqueness as well as its instability are always recurrent difficulties. Additionally, the efficiency and the robustness of the adopted strategy in handling incomplete and inevitably noisy data have to be tested for the case at hand. Due to their complexity and in view of recent technological developments achieved in computer sciences, the popularity of innovative soft computing-based techniques for systems identification in mechanical applications has increased in the last decade. In this context, significant results have been achieved especially using neural networks and genetic algorithms: nowadays, non-classical methodologies based on biological, social or physical paradigms are sufficiently consolidated in both theoretical and practical applications.

What is the role and what are the results actually available about genetic algorithms-based identification for mechanical problems? What are the criticisms and the unexplored potentialities in these applications? Looking for appropriate responses to these questions, we wish to illustrate the state-of-the-art on this very interesting topic.

The chapter is structured as follows. Firstly, we recognize the applications comprising the actual state-of-the-practice in aeronautic, aerospace, civil, mechanical and civil engineering. Covered areas included in this analysis are geomechanic, mechatronic, mechanics of solids and materials science. Subsequently, each part of the genetic algorithms (objective functions, chromosome encoding, initialization of the population, fitness function, selection, crossover, mutation and migration operators, stop criteria) are investigated and apposite references are given. In the third section, we discuss additional strategies that are usually included in the traditional structure of the genetic algorithms. In this sense, modified genetic algorithms based on hybrid numerical strategies and other artifices regarding the population size, the data length and the reduction of the search space are taken into account. Finally, general remarks and suggestions are presented as well as further needed works in this area.

purchase the full-text of this chapter (price £20)

go to the previous chapter
go to the next chapter
return to the table of contents
return to the book description
purchase this book (price £88 +P&P)