Identification of sources through a probabilistic approach: application to vehicles in real usage conditions

The problem of identifying the excitation forces applying on a dynamical system in real usage conditions is addressed and illustrated in the automotive context. The mechanical system is defined jointly with the measurement protocol needed for the "in situ" load identification. A probabilistic way of dealing with the inverse problem intrinsic to the "in situ" approach is proposed: not only the Bayesian context performs a non-supervised regularization of the problem, but it also yields the variance of the estimated loads which is useful to judge the quality of the inversion.