Application of neuronetwork models for solving tasks of durability prediction of corrodible beam constructions

User Rating:  / 0
PoorBest 

Authors:

D.G. Zelentsov, Dr. Sci. (Tech.), Professor, Ukrainian State University of Chemical Technology, Head of the Department of Information Technology, Dnipropetrovsk, Ukraine

A.P. Ivanova, Cand. Sci. (Tech.), Associate Professor, State Higher Educational Institution “National Mining University”, Senior Lecturer of the Department of Construction, Geotechnics and Geomechanics, Dnipropetrovsk, Ukraine

Abstract:

Purpose. To design new models of the deformation process of corrosive beam structural systems with variable geometry characteristics and effective numerical algorithms to solve problems of their stress-strain state and longevity.

Methodology. Reasons resulting in the emergency destruction of multiple-element metal structures were analyzed. Research methods and design methods involving material defectiveness were considered. The ways to improve their efficiency were proposed based on the analysis. The simulation of beam structural components behavior affected by of quasistatic, cyclic, and random loads operating in an aggressive environment forms the basis of the research. The special attention was paid to the influence of different types of deformation on the corrosive wear of such structural components as well as the possibility of their safe exploitation.

Findings. Application of new models of corrosive surface allowed us to design the modified beam finite component with variable rigidity. It will allow us to take into account the change in section shape caused by an aggressive environment and to decrease both the problem order of modified finite element (MFE) and the order of differential equation system (DES) describing corrosive process. Expressions for rigidity matrix components such as finite elements (FE) cannot depend on cross-section type if it is possible to represent it as the total of rectangular fragments. To improve the efficiency of computational algorithm and simplify its logics, it is expedient to determine the type of active constraint before solving the problem of longevity computation. We propose to use two neural networks: to determine the constraint type (neural network 1), and parameters of DES numerical solution for each of the constraints (neural network 2).

Based on the analysis of the factors that affect both the type of active constraint and the accuracy of DES solution, we have assumed the following input network parameters: rolled section number, corrosion rate, and initial stress σ0.

The constraint type depends on the neural network with threshold function of output element activation. The rational value of numerical solution parameters depends on a network with sigmoid activation function.

Neural networks were controlled by the genetic algorithm and algorithm of back propagation of error.

Originality. The model of a corrosive surface has been improved. It was used for the development of rigidity matrix for modified variable-rigidity beam finite element. A modified variable-rigidity beam FE has been designed. The rational parameters of numerical procedures have been developed using neuron networks. This may help us to obtain the solution with an error being less than the maximum admissible value with minimum computational costs.

Practical value. The research contributes to the improvement of the efficiency of solving stress-strain state problems and determination of the longevity of corrosive beam metal structural systems.

References:

1. Расторгуев Б.С. Обеспечение живучести зданий при особых динамических воздействиях / Б.С. Расторгуев // Сейсмостойкое cтроительство. Безопасность cооружений. – 2003. – №4. – С. 45–48.

Rastorguev, B.S. (2003), “Ensuring the survivability of buildings with special dynamic effects”, Seysmostoykoye Stroitelstvo. Bezopasnost Sooruzheniy, no.4, pp. 45–48.

2. Райзер В.Д. Теория надежности в строительном проектировании / Райзер В.Д. – М.: АСВ, 1998. – 304 с.

Rayzer, V.D. (1998), Teoriya nadezhnosti v stroitelnom proektirovanii [A Theory of Reliability in Building Engineering], ASV, Moscow, Russia.

3. Овчинников И. Г. Банк математических моделей коррозионного износа, применяемых для прогнозирования поведения металлоконструкций / И.Г. Овчинников, М.С. Дворкин, Х.А. Сабитов // Проблемы прочности материалов и конструкций, взаимодействующих с агрессивными средами – Саратов: СГТУ, 1993. – С. 141–150.

Ovchinnikov, I.G., Dvorkin, M.S. and Sabitov, Kh.A. (1993), “Bank of mathematical models of corrosive wear, applied for forecasting of metallic constructions behavior”, Problemy Prochnosti Materialov i Konstruktsyy, Vzaimodeystvuyushchikh s Agressivnymi Sredami, SGTU, pp. 141–150.

4. Зеленцов Д.Г. Балочный конечный элемент переменной жесткости для расчета конструкций, подверженных коррозионному износу / Д.Г. Зеленцов // Системні технологии. – Дніпропет­ровськ, 2001. – Вип. 6(17). – С. 64–70.

Zelentsov, D.G. (2001), “Beam final element of variable inflexibility for the calculation of constructions subject to the corrosive wear”, Systemni Tekhnolohii, Dnipropetrovsk, no. 6(17), pp. 64–70.

 

Files:
2015_04_zelentsov
Date 2015-11-03 Filesize 600.81 KB Download 850

Visitors

6237067
Today
This Month
All days
1521
63744
6237067

Guest Book

If you have questions, comments or suggestions, you can write them in our "Guest Book"

Registration data

ISSN (print) 2071-2227,
ISSN (online) 2223-2362.
Journal was registered by Ministry of Justice of Ukraine.
Registration number КВ No.17742-6592PR dated April 27, 2011.

Contacts

D.Yavornytskyi ave.,19, pavilion 3, room 24-а, Dnipro, 49005
Tel.: +38 (056) 746 32 79.
e-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.
You are here: Home Home EngCat Archive 2015 Contents No.4 2015 Geotechnical and mining mechanical engineering, machine building Application of neuronetwork models for solving tasks of durability prediction of corrodible beam constructions