Accuracy modelling of powder metallurgy process using backpropagation neural networks

In the present paper a neural network approach to accurate modelling of the PM process, particularly the production of self-lubricating bearings, is derived. The model is based on a three layer neural network with a backpropagation learning algorithm. In applying the derived model, the deviations in sintered part dimensions are decreased, thus eliminating the need for additional operations to achieve the required accuracy of the final parts. The simulated results demonstrated that the neural network model is more accurate than the standard design procedure based on the statistical processing of experimental data. Also, the neural network exhibits the very useful feature that the same algorithm (and/or configuration) can be used for resolving different tasks (only new training set should be applied).