Topological data analysis for industrial condition monitoring
Condition based monitoring for process and equipment health via topology-derived indicators.
Condition monitoring and predictive maintenance help prevent failures and reduce downtime. In this work, topological data analysis (TDA) was combined with classical (Fourier) signal analysis for condition-based monitoring.
In the first study, TDA and persistent homology were used for signal analysis on offshore sensors and condition monitoring of multiphase flow. The goal was to identify and classify the transition from regular to severe slugging flow using machine learning.
Published in Digital Chemical Engineering, Vol. 4, 100045 (2022). A corrected PDF (formulae typos fixed) is available here.
Then, a work on wind turbine gearbox CBM was presented at PHM Europe 2024. Acceleration and spectra were collected from a wind park in Norway; TDA was applied to point clouds built from vibration time series to derive topology-based health indicators, which were evaluated for fault detection and classification. The paper appears in the PHM Society European Conference proceedings (DOI 10.36001/phme.2024.v8i1.4117); an extended version is also on arXiv.
