In the mechanical transmission system, the bearing is one of the most widely used transmission components. The failure of the bearing will cause serious accident and huge economic loss. Therefore, the remaining life prediction of the bearing is greatly important. In order to predict the remaining life of the bearing, a prediction method combining macro-micro attention, long-short-term memory neural network and isometric mapping is proposed. First, some typical time-domain and frequency-domain characteristics of the vibration signal are calculated .respectively, such as the maximum value, the absolute mean value, the standard deviation, the kurtosis and so on. Then, the principal component of these characteristics is extracted by the isometric mapping method. The importance of fusional characteristic information is filtered via a proposed macro-micro attention mechanism, so that the input weight of neural network data and recursive data can reach multi-level real-time amplification. With the new long short-term memory neural network, the characteristic of the bearing vibration signal can be predicted based on the known fusional characteristic. The experimental results show that the method can predict the remaining life of the bearing well and has higher prediction accuracy than the conventional LSTMs.

Key words: remaining life prediction; long short-term memory neural network; macro-micro attention mechanisms; vibration signal; feature fusion