<ul><li>A novel unsupervised framework was developed for constructing a dynamic Health Indicator (HI) for analyzing rolling bearings' degradation and predicting future trends.</li><li>The framework includes a degradation feature learning module that uses a skip-connection-based autoencoder to extract essential features from raw signals without expert knowledge.</li><li>A new HI-generating module with an inner HI-prediction block captures temporal dependencies between past and current HI states, ensuring effective degradation trend representation and prognostics.</li><li>Experiment results on two bearing lifecycle datasets demonstrate the superiority of the proposed dynamic HI construction method for prognostic tasks compared to other methods.</li></ul>

An Unsupervised Framework for Dynamic Health Indicator Construction and Its Application in Rolling Bearing Prognostics

Discover more