Predicting and Forecasting Fatigue Damage of Automotive Components
Improved remaining useful life (RUL) prediction leveraging fleet data and AI
The traditional approach to vehicle reliability engineering and RUL prediction consists of mileage accumulation or testing under constraints like representative
loads, combined with failure analysis of automotive components. These methods are based on damage accumulation, which are mostly semi-empirical approaches. It can take years to collect enough
history of vehicles where a specific component has actually failed. In the meantime, hundreds of thousands of vehicles incorporating this component have rolled off the production lines. Earlier and more precise RUL prediction of
automotive components could provide a significant competitive advantage to both vehicle manufacturers and TIER 1 suppliers.
Our approach considers additional factors like real-life vehicle usage patterns and environmental conditions out of fleet data right from the start, to provide an improved RUL prediction as early as possible.
The article describes how the dataset we used was generated, using artificial but realistic trip histories. It also shows how our methodology manages with an iterative approach to overcome the issue of “not having enough damaged cars”. A “simple damage” was calculated according to state-of-the-art methods, completed by an “actual damage” based on usage and environmental factors. Based on the estimated effective damage an additional forecasting pipeline allows then the estimation of the potential point of failure, so that improved RUL prediction is made possible.
Read the full article on Medium.com
Get in touch with us to discuss about your application !
+49 (0)6151 3844614
64283 Darmstadt, Germany