Makine Öğrenmesi Metotları Kullanarak Krom III Kaplama Banyosunun Örtme Gücünün Tahmin Edilmesi

Bu çalışmada krom kaplamanın örtme gücünü tahmin etmek ve örtme gücüne etki eden öznitelikleri belirlemek için makine öğrenmesi algoritmaları kullanılmıştır. Bu amaçla GP (Gaussian Process), KNN (K-Nearest Neighbors), RF (Random Forest), SVR (Support Vector Regressor) ve XGB (eXtreme Gradient Boosting) algoritmaları seçilmiş ve bu algoritmaların hiper parametreleri optimize edilmiştir. En yüksek R2 ve en düşük MSE değerlerini veren şartlar belirlenmiştir. Çapraz doğrulama için LOO (Leave-One-Out) metodu kullanılmıştır. En iyi sonuç, SVR metodu ile elde edilmiştir. R2, MSE ve MAPE değeri sırasıyla 0,80, 0,26 ve 18.29 dur. Kaplamanın örtme gücüne etki eden en önemli iki öznitelik borik asit (H3BO3) ve A kimyasalıdır. Bu kimyasalların yüksek seviyeleri kaplamanın örtme gücünü artırmıştır. Tüm algoritmaların hiper parametreleri ızgara tarama yöntemi ile 2 veya daha fazla seviyede optimize edilmiştir. SVR metodunda en etkin iki hiper parametre kernel ve C parametresidir. Kernel ve C hiper parametreleri sırasıyla “rbf” ve 1 olduğu durumda en yüksek R2 değeri elde edilmiştir. Bu çalışma makine öğrenmesi algoritmalarını elektrokaplama sahasına uygulayan ilk çalışmalardandır. Bu yönüyle öncü olma niteliği taşımaktadır.

Prediction of Covering Power of Chromium III Plating Bath Using Machine Learning Methods

In this study, machine learning algorithms were used to predict the covering power of chrome plating and to determine the features that affect the covering power. For this purpose, GP (Gaussian Process), KNN (K-Nearest Neighbors), RF (Random Forest), SVR (Support Vector Regressor) and XGB (eXtreme Gradient Boosting) algorithms were selected and their hyper-parameters were optimized. Conditions giving the highest R2 and lowest MSE values were acquired. The LOO (Leave One Out) method was used for cross validation. The best results were obtained using the SVR method. The R2, MSE and MAPE scores are 0.80, 0.26 and 18.29, respectively. The two most important features affecting the the covering power of the coating are boric acid (H3BO3) and chemical A. High levels of these chemicals increased the covering power of the coating. The hyper-parameters of all algorithms were optimized at 2 or more levels by the grid scan method. The two most effective hyper-parameters in the SVR method are the kernel and the C parameters. The highest R2 value was obtained when the Kernel and C hyper-parameters were “rbf” and 1, respectively. This study is one of the first studies to apply machine learning algorithms to the electroplating field.

PDF

___

[1] Z. Zeng, Y. Sun, J. Zhang, The electrochemical reduction mechanism of trivalent chromium in the presence of formic acid, Electrochem. Commun. 11 (2009) 331–334. https://doi.org/10.1016/J.ELECOM.2008.11.055.
[2] S. Surviliene, O. Nivinskiene, A. Češuniene, A. Selskis, Effect of Cr(III) solution chemistry on electrodeposition of chromium, J. Appl. Electrochem. 36 (2006) 649–654. https://doi.org/10.1007/s10800-005-9105-8.
[3] J.-Y. Lee, M. Kim, S.-C. Kwon, Effect of polyethylene glycol on electrochemically deposited trivalent chromium layers, Trans. Nonferrous Met. Soc. China. 19 (2009) 819–823. https://doi.org/10.1016/S1003-6326(08)60357-X.
[4] S.L. Handy, C.F. Oduoza, T. Pearson, Theoretical aspects of electrodeposition of decorative chromium from trivalent electrolytes and corrosion rate study of different nickel/chromium coatings, Trans. Inst. Met. Finish. 84 (2006) 300– 308. https://doi.org/10.1179/174591906X162946.
[5] X. Ren, Y. Song, A. Liu, J. Zhang, G. Yuan, P. Yang, J. Zhang, M. An, D. Matera, G. Wu, Computational Chemistry and Electrochemical Studies of Adsorption Behavior of Organic Additives during Gold Deposition in Cyanide-free Electrolytes, Electrochim. Acta. 176 (2015) 10–17. https://doi.org/10.1016/j.electacta.2015.06.147.
[6] Z. Feng, A. Liu, L. Ren, J. Zhang, P. Yang, M. An, Computational Chemistry and Electrochemical Mechanism Studies of Auxiliary Complexing Agents Used for Zn-Ni Electroplating in the 5-5’-Diethylhydantoin Electrolyte, J. Electrochem. Soc. 163 (2016) D764–D773. https://doi.org/10.1149/2.0591614jes.
[7] R. Katırcı, E. Sezer, B. Ustamehmetoğlu, Statistical optimisation of organic additives for maximum brightness and brightener analysis in a nickel electroplating bath, Trans. IMF. 93 (2015) 89–96. https://doi.org/10.1179/0020296714Z.000000000219.
[8] K. Sasaki, G.C. Kabushiki, C. Abstracts, P.E.E. Waddell, United States Patent [ 191, (1976) 575–585.
[9] R. Katırcı, A chrome coating from a trivalent chromium bath containing extremely low concentration of Cr³⁺ ions, Int. J. Surf. Sci. Eng. 10 (2016) 73. https://doi.org/10.1504/IJSURFSE.2016.075318.
[10] H.B. Muralidhara, Y. Arthoba Naik, Electrochemical deposition of nanocrystalline zinc on steel substrate from acid zincate bath, Surf. Coatings Technol. 202 (2008) 3403–3412. https://doi.org/10.1016/j.surfcoat.2007.12.012.
[11] B. Lenz, H. Hasselbruch, H. Großmann, A. Mehner, Application of CNN networks for an automatic determination of critical loads in scratch tests on a-C:H:W coatings, Surf. Coatings Technol. 393 (2020) 125764. https://doi.org/10.1016/j.surfcoat.2020.125764.
[12] A. Krizhevsky, I. Sutskever, G.E. Hinton, ImageNet classification with deep convolutional neural networks, Commun. ACM. 60 (2017) 84–90. https://doi.org/10.1145/3065386.
[13] C. Szegedy, V. Vanhoucke, S. Ioffe, J. Shlens, Z. Wojna, Rethinking the Inception Architecture for Computer Vision, Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. 2016-Decem (2016) 2818–2826. https://doi.org/10.1109/CVPR.2016.308.
[14] K. He, X. Zhang, S. Ren, J. Sun, Deep residual learning for image recognition, Proc. IEEE Comput. Soc. Conf. Comput. Vis. Pattern Recognit. 2016-Decem (2016) 770–778. https://doi.org/10.1109/CVPR.2016.90.
[15] K. Simonyan, A. Zisserman, Very deep convolutional networks for large-scale image recognition, 3rd Int. Conf. Learn. Represent. ICLR 2015 - Conf. Track Proc. (2015) 1–14.
[16] J. Zhu, X. Wang, L. Kou, L. Zheng, H. Zhang, Prediction of control parameters corresponding to in-flight particles in atmospheric plasma spray employing convolutional neural networks, Surf. Coatings Technol. 394 (2020) 125862. https://doi.org/10.1016/j.surfcoat.2020.125862.
[17] R. Katirci, H. Aktas, M. Zontul, The prediction of the ZnNi thickness and Ni % of ZnNi alloy electroplating using a machine learning method, Trans. Inst. Met. Finish. 99 (2021) 162–168. https://doi.org/10.1080/00202967.2021.1898183.
[18] C.E. Rasmussen, C.K.I. Williams, Gaussian Processes for Machine Learning (Adaptive Computation and Machine Learning), The MIT Press, 2005.
[19] L. Peterson, K-nearest neighbor, Scholarpedia. 4 (2009) 1883. https://doi.org/10.4249/scholarpedia.1883.
[20] Y.L. Pavlov, Random forests, Random For. (2019) 1–122. https://doi.org/10.1201/9780429469275-8.
[21] C. Cortes, V. Vapnik, Support-vector networks, Mach. Learn. 20 (1995) 273–297. https://doi.org/10.1007/BF00994018.
[22] T. Chen, C. Guestrin, XGBoost, in: Proc. 22nd ACM SIGKDD Int. Conf. Knowl. Discov. Data Min., ACM, New York, NY, USA, 2016: pp. 785–794. https://doi.org/10.1145/2939672.2939785.
[23] G.I. Webb, C. Sammut, C. Perlich, T. Horváth, S. Wrobel, K.B. Korb, W.S. Noble, C. Leslie, M.G. Lagoudakis, N. Quadrianto, W.L. Buntine, N. Quadrianto, W.L. Buntine, L. Getoor, G. Namata, L. Getoor, J. Han, Xin Jin, J.-A. Ting, S. Vijayakumar, S. Schaal, L. De Raedt, Leave-One-Out Cross-Validation, in: Encycl. Mach. Learn., Springer US, Boston, MA, 2011: pp. 600–601. https://doi.org/10.1007/978-0-387-30164-8_469.
[24] C.W. Holland, D.W. Cravens, Fractional Factorial Experimental Designs in Marketing Research, J. Mark. Res. 10 (1973) 270. https://doi.org/10.2307/3149694.
[25] C. Qi, J. Diao, L. Qiu, IEEE Access, 2019, DOI:10.1109/ACCESS.2019.2892062.
[26] J. Caicedo-Acosta, G. A. Castaño, C. Acosta-Medina, A. Alvarez-Meza, G. Castellanos-Dominguez, Sensors, 2021, DOI:10.3390/s21061932.
[27] X. Yang, L. Li, Q. Tao, W. Lu, M. Li, Comput. Mater. Sci., 2021, DOI:10.1016/j.commatsci.2021.110528.
[28] G. C. Cawley, "Leave-One-Out Cross-Validation Based Model Selection Criteria for Weighted LS-SVMs," The 2006 IEEE International Joint Conference on Neural Network Proceedings, 2006, pp. 1661-1668, doi: 10.1109/IJCNN.2006.246634.
[29] Hawkins, D. M. (2004). The Problem of Overfitting. Journal of Chemical Information and Computer Sciences, 44(1), 1–12. https://doi.org/10.1021/ci0342472
[30] Qi, C., Diao, J., & Qiu, L. (2019). On Estimating Model in Feature Selection with Cross-Validation. IEEE Access, 7, 33454–33463. https://doi.org/10.1109/ACCESS.2019.2892062
[31] Tan, X., Bi, W., Hou, X., & Wang, W. (2011). Reliability analysis using radial basis function networks and support vector machines. Computers and Geotechnics, 38(2), 178–186. https://doi.org/10.1016/j.compgeo.2010.11.002.