General contact point response estimation approach for vehicle scanning via Bayesian filtering and smoothing

Contact point response signals offer competitive advantages for bridge modal identification in vehicle scanning. This study proposes a general framework to estimate all time derivatives (displacement, velocity, and acceleration) of the contact point response across different vehicle types, bridge types, and contact models. The proposed framework consists of (i) a novel state-space model that links contact point response to the flexible modal response of the vehicle while accommodating various vehicle measurement scenarios, and (ii) a joint input-state estimation algorithm (via an Augmented Kalman Filter (AKF) and a Rauch-Tung-Striebel smoother). In this context, the unknown input accounts for the contact point response with a kinematic relationship instead of the conventional random walk model. Subsequently, the residual contact point response is obtained by post-processing the contact point response. Using first numerical simulations, the study demonstrates the applicability of the proposed approach and investigates the influence of contact model, vehicle speed, bridge and vehicle types, road roughness, sensor noise, and modeling errors on estimation accuracy. Then, the study further validates the proposed framework by means of a laboratory experiment.