Radul O. Durability computation and optimization of corroding structures with the use of neural network models

Object of study - the stress-strain state of structural systems in corrosive wear. The aim of the thesis: development of effective models, methods and algorithms for durability estimation and optimization of corrosive structures that involve obtaining a solution with the required accuracy. Research methods: stress-strain state of structures was analyzed by finite element method. For optimization the random search method was used in conjunction with the method of rolling admission. The numerical integration of systems of differential equations describing the corrosion process was carried out according to the algorithms, which are based on Euler's method. To construct the approximation model artificial neural networks were used. Scientific novelty: for the first time the value of the required accuracy of the solution of the task of corroding structural optimization was introduced as an additional constraint, for the first time the module of control of solution's accuracy of systems of differential equations, where the neural network models. They were used to determine the integration step, which provides the necessary accuracy in calculating the constraint function with minimal computational cost; further developed the idea of using the maximum permissible error of calculating the constraint function as a criterion rolling admission. To modify a rolling admission for the first time a series of neural networks that were configured to different accuracy solutions of systems of differential equations was used; improved scheme for solving optimization problem with two consecutive single-circuit connections through its more accurate identification using neural network models; improved simulation model for determining longevity stressed corroding plate with a hole through its more accurate identification using neural networks. Theoretical and practical results: a new method for calculating durability and optimizing of corroding structures. In contrast to existing methods, it allows to obtain the solution of tasks with the required accuracy without additional calculations, thus expanding the range of applications of this technique to more complex designs that require high computational costs. The software is developed for definition of durability and optimization, which can be used in scientific research organizations in the study and design of corroding structures. The degree of implementation: methodology and software embedded in the educational process in SHEI "Ukrainian State University of Chemical Technology" and in the design practice of OJSC "Dnepropolimermash". Application: estimation of durability and optimization of corroding structures.