In this study, it is aimed to track the aging trend of the incipient bearing damage in an induction motor which is subjected to an accelerated aging process. For this purpose, a new Multilayer perceptron (MLP) neural network approach is introduced. The input signals are extracted from power spectral densities (PSD) of the vibration signals taken from a 5-HP induction motor. Principal component analysis (PCA) has been applied to select the best possible feature vectors as a dimensionality reduction purpose. Variance and entropy values are used as the targets of the MLP network. The healthy motor condition was modelled by the MLP network considering all load conditions. The results showed that the incipient bearing damage was clearly extracted by the oscillations of the MLP output error.