The criticality of spare parts evaluating model using artificial neural network approach

Abstract

This paper presents artificial neural networks (ANNs) for the criticality evaluating of spare parts in a power plant. Two learning methods were utilized in the ANNs, namely back propagation and genetic algorithms. The reliability of the models was tested by comparing their classification ability with a hold-out sample and an external data set. The results showed that both ANN models had high predictive accuracy. The results also indicate that there was no significant difference between the two learning methods. The proposed ANNs was successful in decreasing inventories holding costs significantly by modifying the unreasonable target service level setting which is confirmed by the corresponding criticality class in the organization. © Springer-Verlag Berlin Heidelberg 2006.