Modelling of the cutting tool stresses in machining of Inconel 718 using artificial neural networks


EXPERT SYSTEMS WITH APPLICATIONS, cilt.36, sa.6, ss.9645-9657, 2009 (SCI İndekslerine Giren Dergi) identifier identifier

  • Cilt numarası: 36 Konu: 6
  • Basım Tarihi: 2009
  • Doi Numarası: 10.1016/j.eswa.2008.12.054
  • Sayfa Sayıları: ss.9645-9657


This study covers two main subjects: (i) The experimental and theoretical analysis: the Cutting forces and indirectly cutting tool stresses, affecting the cutting tool life during machining in metal Cutting, are One Of very important parameters to be necessarily known to select the economical cutting conditions and to Mount the workpiece on machine tools securely. In this paper, the cutting tool stresses (normal, shear and von Mises) in machining of nickel-based Super alloy Inconel 718 have been investigated in respect of the variations in the cutting parameters (cutting speed, feed rate and depth of cut). The cutting forces were measured by a series of experimental measurements and the stress distributions on the cutting tool were analysed by means of the finite element method (FEM) using ANSYS software. ANSYS stress results showed that in point of the cutting tool wear, especially from von Mises stress distributions, the ceramic cutting insert may be possible worn at the distance equal to the depth of cut on the base cutting edge of the cutting tool. Thence, this wear mode will be almost such as the notch wear, and the flank wear on the base cutting edge and grooves in relief face. In terms of the cost of the process of machining, the cutting speed and the feed rate values Must be chosen between 225 and 400 m/min, and 0.1 and 0, 125 mm/rev, respectively. (ii) The mathematical modelling analysis: the use of artificial neural network (ANN) has been proposed to determine the cutting tool Stresses in machining of Inconel 718 as analytic formulas based on working parameters. The best fitting set was obtained with ten neurons in the hidden-layer using back propagation algorithm. After training, it was found the R(2) values are closely 1. (C) 2008 Elsevier Ltd. All rights reserved.