Identify Steel Ball Surface Defect Based on Combination of Dynamic and Static RBF Neural Network

Abstract:

Steel ball, as a rolling body of all kinds of bearings, it direct affects the bearings precision, dynamic performance and service life. This paper introduces the digital image technology Radial Basis Function (RBF)-Neural network, based on extracting the Steel Ball surface defect image features, used the strategy which is combined with static- dynamic clustering to union the two-stage study and design the hidden layer structure. Simulation and experiment show that the RBF-neural network runs stably, has fast convergence and overall accuracy rate of 96%. These can meet the needs of practical application.

• Periodical: Applied Mechanics and Materials (Volumes 16-19)
• Pages: 1000-1004
• DOI: https://doi.org/10.4028/www.scientific.net/AMM.16-19.1000
• Online since: October 2009
• Authors: Yan Ling Zhao, Feng Ling Wu, Peng Wang, Jian Yi Zhang
• Keywords: Combination of Dynamic-Static Clustering, Feature Extraction, RBF Neural Network, Steel Ball Surface Defect

References

[1] U. Zekkeriva and G. Cunevt: IEEE Trans Neural Networks, Vol. 16 (2000), p.851.

[2] X. Ming and D.L. Zhang: Journal of Anhui University (Natural Science Edition), Vol. 24 (2000), p.72. (In Chinese).

[3] Y.F. Sun, C.Y. Liang and Q.F. Meng: Journal of Jilin University (Information Science), Vol. 1 (2002), p.63. (In Chinese).

[4] A.K. Jain, J. Mao and K.M. Mohiuddin: IEEE Computer, Vol. 29 (1996), p.31.

[5] G.B. H. and P. Saratchandran: IEEE Transactions on Systems, Man, and Cybernetics-Part B: Cybernetics, Vol. 34 (2004), p.2284.