Visense for Vehicle Load Estimation and Fatigue Analysis

Fatigue or strain damage is a crucial factor in mechanical component design during the vehicle development process. Especially for special-purpose vehicles like parcel delivery vehicles or taxis, knowledge about degraded components induced by specific usages can be enormously helpful during component layout. Usually, fatigue or strain damage is derived from load history data using the so-called rain flow algorithm. However, precise measurement of the load history on vehicle components using special sensors can be cost-intensive or limited for technical reasons. To overcome these obstacles, we propose a data-driven neural network model, ViSense, to accurately estimate load behavior on vehicle components based on standard vehicle sensor data. Furthermore, we show that Visense reliably outperforms standard neural network models in deriving the mechanical fatigue damage, providing a precise and cost-effective alternative to hardware sensors.