Application of neural networks in predicting the quality of machining of carrying composite structures
Design, construction and manufacturing of flying vehicles
Аuthors1*, 2**, 2***
1. Southern Scientific Center of the Russian Academy of Sciences, 41, Chekhov str., Rostov-on-Don, 344006, Russia
2. Don State Technical University, DSTU, 1, Gagarin square, Rostov-on-Don, 344003, Russia
Manufacturing of new aircrafts requires development and improvement of new materials and technologies, ensuring ever-increasing demands for quality and operation reliability. Most carrying aircraft structures made of polymeric composites are machined to produce the high quality holes providing reliable fixation and assembly of composite structure (such as, spars of the main and tail helicopter rotors blades). Delamination that can reduce the material structural integrity is basic technological problem, occurring while fiber reinforced composite materials drilling. Tool’s geometry and machining conditions are the most important factors affecting the quality of the processed holes.
The main purpose of this work consisted in the study the process of drilling holes in a multilayered polymeric composite of “GFRP-titanium” type, which is employed in the design of the Mi-28 helicopter rotor blades spar, as well as for soft tool predicting the quality of the holes development.
Using a series of experimental data, including the delamination sizes depending on the machining conditions, the database for the further modeling has been created in a matrix form. Our first attempt to build an empirical description of delamination sizes, depending on two design variables (cutting speed and drilling feed) has been unsuccessful, probably, due to nonlinearity, generic for the drilling process.
To overcome this difficulty we proposed and tested an approach, based on the artificial neural networks (ANN) implementation, to predict the quality of the holes drilled in the titanium foil reinforced glass fiber epoxy-based plastics. ANN training has been conducted using an errors back-propagation algorithm. Testing of the prediction accuracy revealed that ANN could provide the quality of size prediction exceeding 97%.
According to the obtained results, we recommend to use the artificial neural network for the delamination at the GFRP drilling holes size prediction. In the process of statistics accumulation ANN are able to carry out self-learning and produce the results based on the newly obtained information, and, thus, adequately predict the quality of machined holes.
Keywords:drilling glass fiber reinforced plastic (GFRP), delaminations, artificial neural networks, method of conjugate gradients
Efimov E.N., Shevgunov T.Ya. Trudy MAI, 2012, no.51: http://www.mai.ru/science/trudy/published.php?ID=29159
Grischenko S.V. Trudy MAI, 2015, no. 84: http://www.mai.ru/science/trudy/published.php?ID=63011
Ivanov Yu.N. Izvestiya Samarskogo nauchnogo tsentra RAN, 2014, vol. 16, no. 1(5). pp. 1402–1406.
Medvedev V.S., Potemkin V.G. Neironnye seti. MATLAB 6. (Neural networks. MATLAB 6.), Moscow, DIALOG-MIFI, 2002, 496 p.
Panteleev A.V., Letova T.A. Metody optimizatsii v primerakh i zadachakh (Optimization methods in examples and problems), Moscow, Vysshaya shkola, 2005, 544 p.
Altinkok N., Koker R. Modelling of the prediction of tensile and density properties in particle reinforced metal matrix composites by using neural networks. Materials and Design. 27. 2006. pp. 625–631.
Chakraborty D. Artificial neural network based delamination prediction in laminated composites. Materials and Design. 26. 2005. pp. 1–7.
Gaitonde V.N. A study aimed at minimizing delamination during drilling of CFRP composites. Journal of Composite Materials. 45 (22). 2011. pp. 2359–2368.
Sorrentino L., Turchetta S., Colella C. In-process force and temperature monitoring while drilling composites. JEC composites magazine. № 102. January-February 2016. pp. 46-48.
Mishra R., Malik J, Singh I. Singh Prediction of drilling-induced damage in unidirectional glass-fibre-reinforced plastic laminates using an artificial neural network. Proceedings of the Institution of Mechanical Engineers. Part B. Journal of Engineering Manufacture. 2010. P. 733-738.