Title :
Combination predicting neural network model for E4303 electrode mechanical properties
Author :
Huang, Jun ; Xu, Yuelan
Author_Institution :
Dept. of Mater. Sci. & Eng., Nanjing Univ. of Sci. & Technol., Nanjing, China
Abstract :
As for more predicting errors of deposited metal impacting toughness of E4303 structural steel electrode with low alloy, related sample data are acquired by experiments. A nonlinear combination predicting neural network model for E4303 electrode mechanical properties is build based on predicting data acquired by BP, RBF and adaptive fuzzy neural network. To validate the validity of the model, experiment data except training samples are selected to predict with the combination model. Results show that predicting average relative error of impacting toughness is below 5% and it satisfies the practice producing demands. Compared with single model, the accuracy and stability of this combination model is greatly improved.
Keywords :
backpropagation; electrodes; fuzzy neural nets; impact (mechanical); mechanical engineering computing; mechanical properties; metals; steel; BP neural network; E4303 electrode mechanical property; E4303 structural steel electrode; RBF neural network; adaptive fuzzy neural network; average relative error; deposited metal; impacting toughness; nonlinear combination predicting neural network model; predicting error; Artificial neural networks; Carbon; Data models; Electrodes; Mathematical model; Metals; Predictive models; BF neural network; BP neural network; adaptive fuzzy neural network; electrode mechanical properties; nonlinear combination prediction;
Conference_Titel :
Natural Computation (ICNC), 2010 Sixth International Conference on
Print_ISBN :
978-1-4244-5958-2
DOI :
10.1109/ICNC.2010.5583763