Virtual sensing strategies utilizing various model complexities: industrial applications

Virtual sensing is an emerging technology with a wide application potential for the commercial vehicle industry. Through the combination of sensor measurement data with simulation models, the virtual sensing methodology enables the estimation of physical quantities of which the direct measurement is otherwise not feasible, for reasons of e.g. cost or accessible space. The goal of this work is to explore and validate the capabilities of virtual sensing on industrial setups. Three demonstrators are presented, targeting different applications and system (model) complexities. The first demonstrator estimates the dynamic load torque input on an electro-mechanical drivetrain by means of a multi-physical lumped-parameter (1D) model. The second example shows the estimation of the full-field strain response of a twist-beam rear suspension under dynamic load excitation from a reduced sensor set combined with a linear Finite Element (FE) model. The third example demonstrates the estimation of the wheel center forces and moments on a McPherson suspension from strain measurements in combination with a nonlinear, flexible multi-body model.