350 rub
Journal Neurocomputers №9 for 2015 г.
Article in number:
Dynamic neural network model for the process of electrolytic plasma stripping of aluminide coating from nickel superalloy
Keywords:
neural network modeling
neural networks with radial basis functions
multilayer feedback perceptron
electrolytic plasma processing
coatings stripping
Authors:
M.V. Gorbatkov - Post-graduate Student, Department of Theoretical Basis of Electrical Engineering,
Ufa State Aviation Technical University. E-mail: mikesg@mail.ru
E.V. Parfenov - Dr.Sc. (Eng.), Professor, Department of Theoretical Basis of Electrical Engineering, Ufa State Aviation Technical University. E-mail: pev_us@yahoo.com
S.V. Zhernakov - Dr.Sc. (Eng.), Professor,
Head of the Department of Electronics and Biomedical Technologies, Ufa State Aviation Technical University
E-mail: zhsviit@mail.ru
R.R. Nevyantseva - Ph.D. (Chem.), Associate Professor, Department of General Chemistry,
Ufa State Aviation Technical University
E-mail: rrnev@mail.ru
Abstract:
The article proposes a method for designing of a dynamic neural network model for electrolytic plasma processing (EPP) and provides an example for an aluminide coating stripping from a nickel superalloy. Analysis of the experimental data shows that the coating thickness and roughness decrease non-linearly with time; also, the coating stripping rate increases with decreasing voltage and electrolyte temperature. To simulate the process dynamics, the following model swere used: static neural network models based on the experimental data, and a dynamic model designed using computational results from the static model.
The static models of the EPP process have been developed using generalized regression neural network (GRNN) structures, which helped to gradually refine the input variable steps in order to supply a training set for the dynamic model with the target values ΔU = 6.25 B, ΔT = 2.5 °C and Δt = 1 min. To evaluate the adequacy of the models, the probability deviation interval for the coating thickness and roughness values around their mean values was calculated; the approximation quality of the obtained models was evaluated using coefficient of determination.
The dynamic neural network model of the EPP process was implemented using a multi-layer perceptron with feedback (NARX); it has a hidden layer with a time delay. The model is capable of the transient response simulation for the coating thickness and surface roughness with respect to the input voltage and the electrolyte temperature variation. The model can be applied as a reference model in the field of automated process control for electrolytic plasma processing.
Pages: 10-19
References
- Suminov I.V., Belkin P.N., EHpelfeld A.V., Ljudin V.B., Krit B.L., Borisov A.M. Plazmenno-ehlektroliticheskoe modificirovanie poverkhnosti metallov i splavov. M.: Tekhnosfera. 2011. 464 s.
- Parfenov E.V., Nevjanceva R.R., Gorbatkov S.A., Erokhin A.L. EHlektrolitno-plazmennaja obrabotka: modelirovanie, diagnostika, upravlenie. M.: Mashinostroenie. 2014. 380 s.
- Yerokhin A.L., Nie X., Leyland A., Matthews A., Dowey S.J. Plasma electrolysis for surface engineering // Surface and Coatings Technology. 1999. V. 122. P. 73-93.
- Matykina E., Arrabal R., Pardo A., Mohedano M., Mingo B., Rodríguez I., González J. Energy-efficient PEO process of aluminium alloys // Materials Letters. 2014. V. 127. R. 13-16.
- Parfenov E.V., Lazarev D.M., Fatkullin A.R., Erokhin A.L. Matematicheskoe modelirovanie processa ehlektrolitno-plazmennojj obrabotki s ispolzovaniem nejjronnykh setejj // Nejjrokompjutery: razrabotka, primenenie. 2011. № 3. S. 47-55.
- Parfenov E.V., Nevjanceva R.R., Bybin A.A., Mukaeva V.R. Vybor optimalnykh uslovijj udalenija aljuminidnogo pokrytija s poverkhnosti zharoprochnogo nikelevogo splava ehlektrolitno-plazmennym metodom // Fizika i khimija obrabotki materialov. 2010. № 4. S. 19-24.
- Parfenov E.V., Nevjanceva R.R., Bybin A.A., Mukaeva V.R. Vlijanie uslovijj obrabotki na process udalenija zharostojjkogo aljuminidnogo pokrytija ehlektrolitno-plazmennym metodom pri stabilizacii temperatury ehlektrolita // Fizika i khimija obrabotki materialov. 2011. № 6. S. 65-70.
- KHajjkin S.Nejjronnye seti: polnyjj kurs: Per. s angl. M.: Izdatelskijj dom «Viljams». 2006.1104 s.
- Hudson M.B., Hagan M., Demuth B. MATLAB Neural Network Toolbox User-s Guide. MathWorks. 2014.
- ZHernakov S.V., Kobylev S.V. Ispolzovanie rekurrentnykh nejjrosetejj v zadache diagnostiki tekhnicheskogo sostojanija gidromekhanicheskogo avtomata razgona // Vestnik UGATU. 2011. T. 15. № 3 (43). S. 142-156.
- Parfenov E.V., Nevjanceva R.R., Bybin A.A., Parfenov L.V. Programma nejjrosetevogo modelirovanija tekhnologicheskikh processov (na primere ehlektrolitno-plazmennojj obrabotki) / Svidetelstvo o gosudarstvennojj registracii programmy dlja EHVM № 2007615051 / Zareg. 05.12.2007.
- Gorbatkov M.V., Parfenov E.V., ZHernakov S.V., Mukaeva V.R. Programma postroenija dinamicheskikh nejjrosetevykh modelejj processa ehlektrolitno-plazmennojj obrabotki / Svidetelstvo o gosudarstvennojj registracii programmy dlja EHVM № 2014617942 / Zareg. 06.08.2014.