RuAviation

Within the framework of computational and experimental research of flight and aerodynamic performance improvement technologies, structural design methods and development of “state of the art” support technologies for advanced transport category aircraft, the research department of fatigue and static strength of the SibNIA named after S.A. Chaplygin conducted researches on the effectiveness of genetic algorithms in designing bionic structural and force-force schemes for air vehicles.

“SibNIA is investigating the synthesis of bionic structures of the power bracket power linkage assembly of the flying laboratory based on the Yak-40 aircraft based on the method of topological optimization using neural networks,” said the press service of the institute. – Topological optimization is one of the approaches to product design, which provides bionic design, consisting of the introduction of changes to a structure or part with the creation of new boundaries of the body volume and the removal of existing ones, in order to optimize according to criteria of minimizing mass, maximum stiffness or the spectrum of natural frequencies, while maintaining the strength requirements.

Bionic engineering or bionic design involves the creation of products that resemble complex surfaces formed in nature. The artificial neural network is a mathematical model which is based on the principle of organisation and functioning of neural networks of nerve cells in a living organism. The method of neural network topological optimisation models the evolutionary process of a given population (set of vectors) which is multiplied and affected by mutations, resulting in natural selection based on minimisation of the target function. In this paper, a single formal perseptron-type neuron, which calculates the total input signal according to some rule from the whole set of input signals, was used.

The application of neural networks in topological optimization method gives qualitatively similar both in form and weight to optimized models in comparison with classical methods of topological optimization. Nevertheless, the application of neural network algorithm slightly increases the convergence speed of the topological optimization method.

“SibNIA specialists have developed a topological optimization algorithm using neural networks and performed topological optimization of the previously prepared “enlarged” bracket. The research is continuing,” added the institute’s spokesperson.