Journal of Harbin University of Science and Technology (Jun 2017)

Inverse Solutionof BP Neural Network for Laser Ｒemelting Parameters

  • LIU Li-jun,
  • JIANG Ya-qing,
  • WANG Xiao-peng,
  • YAO Ji-rong

DOI

Abstract

Aim at highly nonlinear mapping relationship between the laser processing parameters and the melting cell body’s transverse size，a method of reverse engineering laser melting parameters by back － propagation ( BP) neural network was put forward． The model was constructed by BP neural network，and the prediction error was reduced to less than 3% after training for many times． The DIEVAＲ die steel was melted by reverse engineering laser parameters，and the results show that the error was 1． 33% between the transverse dimensions of the melting cell body and the expected，the expected precision can be met well． Thermal fatigue property of the melted and non － melted DIEVAＲ die steel has been studied． The analysis about cracks growth presents that thermal fatigue property of DIEVAＲ die steel melted by the reverse engineering parameters has been greatly improved． The melting cell body could block crack effectively．

Keywords