Predicting the Mechanical Properties of Electric Porcelain Using ANN and MLR

Abstract:

Electric porcelain has a long history of being used as insulators and bushings in Power T&D. Its microstructure and composition has a large influence on the mechanical performance. In this paper, two different methods, including Artificial Neural Network (ANN) and Multi-Linear Regression (MLR), were employed for predicting mechanical properties of electric porcelain including elastic modulus, bending strength and fracture toughness. Most predicted values by ANN are in accordance with the actual values. MLR gives the best accurate results within the low average error level of ~1%. According to the statistical analysis, two methods are reliable and their predicted results are effective. Formulae obtained from MLR are recommended to be used for predicting the mechanical properties of electric porcelain as experienced formulae.

• Periodical: Advanced Materials Research (Volumes 652-654)
• Pages: 298-303
• DOI: https://doi.org/10.4028/www.scientific.net/AMR.652-654.298
• Online since: January 2013
• Authors: Li Qiao, Lei Chen, Chun Li
• Keywords: Artificial Neural Network (ANN), Electric Porcelain, Mechanical Property, Multi-Linear Regression

References

[1] J. Liebermann: American Ceramic Society Bulletin Vol. 80(6) (2001), pp.37-43.

[2] J. Liebermann, W. Schulle: American ceramic society bulletin Vol. 81(2) (2002), pp.33-38.

[3] Y. S. Li: Process and technology of electric porcelain. Chemical industry press of China, (2007).

[4] L. H. Liu, D. M. Sun: Ceramics Vol. 4 (2005), pp.10-12.

[5] D. M. Sun, L. H. Liu: Materials science & technology Vol. 13(5) (2005), pp.456-458.

[6] C. Z. Huang, L. Zhang: Journal of materials processing technology Vol. 129 (2002), pp.399-402.

[7] http: /marketing. byu. edu/htmlpages/books/pcmds/REGRESS. html.

[8] Q. Y. Xie, J. E. Zhou: Insulators and surge arresters Vol. 175 (2000), pp.12-18.