Prediction Models and Generalization Performance Study in Electrical Discharge Machining

Abstract:

Article Preview

In the past decade, artificial neural network(ANN) has been applied in Electrical discharge machining(EDM). However, most of them only discuss parameter prediction or optimization result, few tell how to improve generalization performance. In this study, machining process models have been established based on different training algorithms of ANN, namely Levenberg-Marquardt algorithm (LM), Resilient algorithm (RP), Scaled Conjugate Gradient algorithm (SCG) and Quasi-Newton algorithm(BFGS). All models have been trained by same experimental data, checked by another group data, their generalization performance are compared. Take LM as the example, some main factors that may influence generalization performance are discussed.

Info:

Periodical:

Edited by:

Kai Cheng, Yingxue Yao and Liang Zhou

Pages:

677-681

Citation:

Q. Gao et al., "Prediction Models and Generalization Performance Study in Electrical Discharge Machining", Applied Mechanics and Materials, Vols. 10-12, pp. 677-681, 2008

Online since:

December 2007

Export:

Price:

$38.00

[1] S.L. Chen, S.F. Hsieh, H.C. Lin and et al: Materials Science and Engineering A, Vol. 445-446 (2007), pp.486-492.

[2] K.M. Tsai, P.J. Wang: International Journal of Machine Tools & Manufacture, Vol. 41 (2001), pp.1385-1403.

[3] F.G. Cao, D.Y. Yang: Journal of Materials Processing Technology, Vol. 149 (2004), pp.83-87.

[4] Shuvra Das, Mathias Klotz and F. Klocke: Journal of Materials Processing Technology, Vol. 142 (2003), pp.434-451.

DOI: https://doi.org/10.1016/s0924-0136(03)00624-1

[5] C.H. Dong: Matlab neural network and application. Peking (National defense industry press, China 2005) (in chinese).

[6] Debabrata. Mandal, Surjya K. Pal and Partha. Saha: Journal of Materials Processing Technology, (2007).