Data-driven prognostics of rolling element bearings using a novel Error Based Evolving Takagi–Sugeno Fuzzy Model

This paper proposes a novel Error Based Evolving Takagi–Sugeno Fuzzy Model (EBeTS) and a new data-driven approach to fault prognostics based on that fuzzy model. The proposed evolving Takagi–Sugeno (TS) model is useful for fault prognostics when the degradation phenomena exhibit nonlinear and time-varying dynamics because the model can represent these characteristics. Since it is an evolving model, it learns the degradation behavior from stream data, although historical data can be used to improve it. Two well-established benchmarks are used to evaluate the EBeTS model and the proposed EBeTS-based prognostics approach. The experiments indicate that the proposed EBeTS-based prognostics approach can take advantage of both historical and new online data to estimate the Remaining Useful Life (RUL) and its uncertainties. Moreover, in most of the cases, it may outperform other methods that do not manage estimation errors and new data incorporation, e.g., fuzzy interacting multiple filters (IMMF) models, Evolving Extended Takagi–Sugeno (exTS) models, and Autoregressive Moving Average (ARMA) model models. •A novel evolving Takagi–Sugeno fuzzy model (called EBeTS) is proposed.•A new evolving prognostics approach using EBeTS is presented.•The proposed prognostic approach incorporates new data from the unit under test instead of using only historical data.•Applications to real bearings degradation from PRONOSTIA platform are discussed.•Comparisons to other approaches illustrate the capability of the method.