Visibility Graphs, Persistent Homology, and Rolling Element Bearing Fault Detection

Topological data analysis (TDA) is fast becoming a pillar in modern machine learning methodologies. Persistent homology is a TDA tool that characterizes the shape of data, which can then be used to train machine learning algorithms. Persistent homology can be sensitive to the choice of metrics, however. We propose a novel method that transforms time series data into visibility graphs, from which persistent homology is applied to extract topological features; we show that this method is effective for fault detection in rolling element bearings.