Collision-prevention systems need an accurate view of brake performance to set safe stopping thresholds. Brake wear is therefore both a maintenance problem and a safety input.
The AFRICOMP 2026 presentation examined a multi-physics prognosis workflow using experimentally generated degradation data. The difficult case was an outlier: wear behaviour unlike the history used to build the initial prediction.
That situation exposes a central weakness of purely historical models. A physical model can provide structure when the observed trajectory changes, while data update the model to the particular asset. The study uses that combination to explore how prognosis can remain useful when a replacement component, assembly difference or operating factor creates unfamiliar degradation.
The downloadable conference deck records the research question, experiment and modelling comparison.