Application of Improved Neural Network in Optimizing Welding Parameter

贾剑平,徐坤刚,李志刚
DOI: https://doi.org/10.3969/j.issn.1001-3814.2008.21.029
2008-01-01
Abstract:The optimal method of L-M(Levenberg-Marquardt) was proposed for improved BP neural network.The relationship between the process parameters and the formation of weld was estabilshed.The predictive model for the rotating arc was put forward,and orthogonal test were arranged using the six factors and two levels to ensure the representative sample of network training.The results indicate that the method has the high predictive precision,and the max relative error does not surpass 3%.Establishment of the network model provides a theoretical basis for optimizing process parameters of rotating arc.
What problem does this paper attempt to address?