Derin öğrenme yöntemi ile temas uzunlukları tahmini

Year 2023, Volume: 13 Issue: 2, 458 - 470, 15.04.2023

Alper Polat

https://doi.org/10.17714/gumusfenbil.1122225

Abstract

Mühendislikte yaygın olarak karşılaşılan problemlerden biri de temas problemleridir. Son yıllarda araştırmacılar karmaşık ve uzun matematiksel ifadeler içeren çözümler yerine daha kısa sürede etkili çözümler sunabilen alternatif yöntemlere yönelmişlerdir . Bu çalışmada, elastik iki dairesel punch ile bastırılan homojen elastik tabakada meydana gelen temas uzunluklarının tahmini yapılmıştır. Bu amaçla makine öğrenmesi alanında son zamanların popüler konusu olan derin öğrenme tekniği kullanılmıştır. Derin öğrenme yapılarından Derin Öğrenme Sinir Ağı (DNN) kullanılarak tahmin için yeni bir model tasarlanmıştır. Homojen elastik tabaka ile punchların kayma modülleri oranı ve punch yarıçaplarından oluşan giriş parametreleri ile beslenen DNN modelinin çıkışında temas uzunluklarının tahmini sağlanmıştır. Modelin eğitimi için analitik çözüm ile elde edilen veriler kullanılmıştır. Ayrıca, sonlu elemanlar yöntemi ile çözümden elde edilen sonuçlar sunulmuş ve DNN sonuçları desteklenmiştir. Çalışmada elde edilen sonuçlar, literatürdeki elastisite teorisi ve klasik Neural Network ile yapılan çözümlerden elde edilen sonuçlarla kıyaslanmıştır. Sonuç olarak klasik Neural Network ile yapılan çözüme kıyasla DNN modeli çok daha kısa sürede ve daha az hatayla sonuçlar elde etmiştir. Sunulan bu modelin temas uzunluğu tahmininde kullanılabilecek etkili bir yaklaşım olduğu söylenebilir.

Keywords

Temas problemi, derin öğrenme, sonlu elemanlar yöntemi, makine öğrenmesi

Estimation of contact lengths using deep neural network

Abstract

One of the most common problems in engineering is contact problems. In recent years, researchers have turned to alternative methods that can offer effective solutions in a shorter time, instead of solutions containing complex and long mathematical expressions. This study focuses on the estimation of the contact lengths in a homogeneous elastic layer suppressed by two elastic punches with two solution methods. Firstly, a new model was designed for estimation using Deep Learning Neural Network (DNN), one of the deep learning structures. Estimation of contact lengths was provided with the output of the DNN model, which was fed with the homogeneous elastic layer, the ratio of shear modules of the punches and the input parameters of punch radii. The finite element method was used as the second solution method. The problem was modeled in the ANSYS programme and the solution was made with the same parameters used in DNN modeled. The results obtained from both solutions were compared with the solutions obtained by the theory of elasticity and classical NN in the literature. It had been seen that the results obtained with DNN and ANSYS were compatible with the results obtained with analytical and classical NN and the margin of error was smaller.

Keywords

contact problem, deep neural network, finite element method, machine learning

References

• Abhilash, M.N., & Murthy, H. (2014). Finite element analysis of 2-D elastic contacts involving FGMs, International Journal of Computer Methods Engineering Science Mechanic, 15(3), 253–7. https://doi.org/10.1080/15502287.2014.882445
• Alinia Y., Aisaee, A. & Hosseini-Nasab, M. (2019). Stress analysis in rolling contact problem of a finite thickness FGM layer, Meccanica, 54, 183–203. https://doi.org/10.1007/s11012-018-00925-w
• ANSYS. (2016). Swanson Analysis Systems Inc., Houston PA, USA.
• Bengio Y., Lamblin P., Popovici D. & Larochelle H. (2006). Greedy layer-wise training of deep networks, In Advances in Neural Information Processing Systems, 19, 153-160.
• Chidlow S.J., Chong W.W.F. & Teodorescu M. (2013). On the two-dimensional solution of both adhesive and non-adhesive contactproblems involving functionally graded materials, European Journal of Mechanics A/Solids. 39, 86–103. https://doi.org/10.1016/j.euromechsol.2012.10.008
• Çakiroğlu E., Çömez I. & Erdöl R. (2011). İki elastik çeyrek düzleme oturan ve dairesel rijit bir punch ile bastırılan elastik tabaka probleminde temas mesafelerinin yapay sinir ağı ile hesabı, VII. Ulusal Mekanik Kongresi (pp. 192-201), Elazığ.
• Çelik Y., Talo, M., Yildirim Ö., Karabatak M. & Acharya U.R. (2020). Automated invasive ductal carcinoma detection based using deep transfer learning with whole-slide images, Pattern Recognition Letters, 133, 232-239. https://doi.org/10.1016/j.patrec.2020.03.011
• El-Borgi S. & Çömez I. (2017). A receding frictional contact problem between a graded layer and a homogeneous substrate pressed by a rigid punch, Mechanics of Materials, 114, 201-214. https://doi.org/10.1016/j.mechmat.2017.08.003
• Goodfellow I., Bengio Y. & Courville A. (2016). Deep learning, T. Dietterich, Ed. London, England: The MIT Press.
• Hinton G. E., Osindero S. & Teh Y. W. (2006). A fast learning algorithm for deep belief nets, Neural Computation. 18(7), 1527-1554.
• Johnson K.L. (1985). Contact mechanics, First Edition, Cambrigde University Press, Cambridge.
• Kahya V., Özşahin T.Ş., Birinci A. & Erdöl R. (2007). A receding contact problem for an anisotropic elastic medium consisting of a layer and a half plane, International Journal of Solids and Structures. 44, 5695-5710.
• Khaleghian S., Ghasemalizadeh O. & Taheri S. (2016). Estimation of the tire contact patch length and normal load using intelligent tires and its application in small ground robot to estimate the tire-road friction, Tire Science and Technology. 44(4), 248-261. https://doi.org/10.2346/tire.16.440402
• Krizhevsky A., Sutskever I. & Hinton G.E. (2015). Imagenet classification with deep convolutional neural networks, In Advances in Neural Information Processing Systems, 25, 1097-1105.
• Lecun Y., Bengio Y. & Hinton G. (2015). Deep learning, Nature, 521(7553), 436-444. https://doi.org/10.1038/nature14539
• Liu Z., Yan J. & Mi C. (2018). On the receding contact between a two-layer inhomogeneous laminate and a half-plane, Structural Engineering and Mechanics, 66(3), 329-341. https://doi.org/10.12989/sem.2018.66.3.329
• Murat F., Yildirim Ö., Talo M., Baloglu U.B., Demir Y. & Acharya U.R. (2020). Application of deep learning techniques for heartbeats detection using ECG signals-Analysis and Review, Computers in Biology and Medicine, 120,103726. https://doi.org/10.1016/j.compbiomed.2020.103726
• Özşahin T.Ş., Birinci A. & Çakıroğlu A.O. (2004). Prediction of contact lengths between an elastic layer and two elastic punches with neural networks, Structural Engineering and Mechanics, 18(4), 441-459.
• Polat A., Kaya Y. & Özsahin T.Ş. (2018). Analytical solution to continuous contact problem for a functionally graded layer loaded through two dissimilar rigid punches, Meccanica, 53(14), 3565-3577. https://doi.org/10.1007/s11012-018-0902-7
• Polat, A. Kaya, Y. Kouider B. & Özşahin T.Ş. (2019). Frictionless contact problem for a functionally graded layer loaded through two rigid punches using finite element method, Journal of Mechanics, 35(5), 591-600. https://doi.org/10.1017/jmech.2018.55
• Rapetto M.P., Almqvist A., Larsson R. & Lugta P.M. (2006). On the influence of surface roughness on real area of contact in normal, dry, friction free, rough contact by using a neural network, Wear, 266(5-6) 592-595.
• Rhimi M., El-Borgi S. & Lajnef N. (2011). A double receding contact axisymmetric problem between a functionally graded layer and a homogeneous substrate, Mechanics of Materials, 43, 787–798. https://doi.org/10.1016/j.mechmat.2011.08.013
• Yaylacı M. (2017). Comparison between numerical and analytical solutions for the receding contact problem, Sigma Journal of Engineering and Natural Sciences, 2, 333-346.
• Yan J. & Mi C. (2017). Double contact analysis of multilayered elastic structures involving functionally graded materials, Archieves of Mechanics, 69(3), 199-221.
• Yıldırım Ö., Pławiak P., Tan R.S. & Acharya U.R. (2018). Arrhythmia detection using deep convolutional neural network with long duration ECG signals, Computers in Biology and Medicine, 102, 411-420. https://doi.org/10.1016/j.compbiomed.2018.09.009