Identification of moving loads in time domain considering uncertainty in the computational model and the boundary conditions

Most existing research on the identification of moving loads in the presence of uncertainties primarily focuses on parametric uncertainties related to the variability or the lack of knowledge of some parameters of the computational model. Such an approach does not allow the consideration of uncertainties related to modelling errors, for instance those related to the discretisation of the structure using the Finite Element (FE) method, simplifying assumptions when deriving the constituent equations, and the idealisation of the boundary conditions. In this paper, a methodology that investigates moving load identification in presence of model-form uncertainties using a non-parametric approach is established. A substructuring-like approach is used to reduce the computational model complexity and to separate the supporting structure from its boundaries, enabling independent control over uncertainty levels within each component of the computational model. A non-parametric probabilistic approach is introduced to the inner part of the reduced-order computational model, while a two-level probabilistic approach is applied to model the uncertainties in the boundaries of the supporting structure. This formulation offers several advantages, notably reducing computational time and accommodating and controlling various types of uncertainties. The efficiency and applicability of this approach are demonstrated through several numerical examples. • A substructuring approach is used to separate the inner part from its boundary. • A general random spring model is introduced to model uncertainties in the boundary. • A generalised- α integration scheme is used to identify the stochastic moving loads. • Statistics of the moving load are calculated using the Monte Carlo method.