Scholarly record

APPLICATION OF AN ARTIFICIAL NEURAL NETWORK FOR PREDICTION OF THE WEAR RESISTANCE OF SINTERED IRON ALLOYS

Florin Marin

Abstract

Neural network method is used to solve complex modeling problems such as classification, estimation and pattern recognition. There are two main categories of ANNs which can be applied whether in regression or classification: the supervised and the unsupervised. In supervised model, all the data is labeled and the network is learned to predict the output values from the input data as previous experimental. In unsupervised model, all the data is unlabeled and the network is learned to inherent structure from the input values. Application of artificial neural networks (ANN) is an efficient solution with applicability in all possible fields due to its robustness and simplicity. In this paper, an apply of ANN for the purpose of predicting the wear behaviour of three different powder metallurgy materials was studied. The materials tested were powders mixtures. The powders were consolidated at a pressures of 400 and 600 MPa. After compaction, the green compacts were sintered at 1150o C for 60 minutes. The various densities of sintered P/M specimens were subjected to dry sliding wear tests. The results of wear tests were compared, analysed and predicted by applying the ANN technique. It is observed that the ANN predicted values are in good agreement with the experimental values obtained. The alloying elements had resulted in enhancing the wear property of the iron based powder due to their carbides formation in the microstructure.

Publication details

Back to publication list