Prediction of deformation characteristics of sintered aluminium preforms using neural networks

N. Selvakumar, P. Radha, R. Narayanasamy, M. J. Davidson

Modelling and Simulation in Materials Science and Engineering · 2004 · 23 citations · 3 references

Concepts

Abstract

Neural networks (NNs) are employed to study the deformation characteristics of sintered aluminium preforms. The proposed NN model has used the measured parameters, namely the load, the aspect ratio and the initial preform fractional density ratio to predict multiple material characteristics, namely the axial stress, the hoop stress, the hydrostatic stress, the axial strain, the hoop strain and the Poisson's ratio. The model is based on a 'four layered NN' with back propagation learning algorithm. The experimental set-up available in the laboratory has been used to get the training data for the sintered aluminium with various preform densities and different aspect ratios (0.50, 0.75 and 1.00) using MoS2 as lubricant. The predicted values from the proposed NN coincide well with the experimental values. In addition, a comparative study between the regression analysis and the NN revealed that the NN can predict the material characteristics of sintered aluminium preform better than regression polynomials within a few per cent error.

References

3