In leading-edge industrial applications, assessing structural integrity is an important aspect of safety requirements. Structural Health Monitoring (SHM) proposes to use sensors and signal processing units in situ. One of the most attractive SHM techniques relies on ultrasonic guided waves. Operational conditions can change wave propagation and therefore affect the interpretation of recorded signals. Modeling and simulation can be helpful tools for the design or the reliability assessment of SHM solutions. In , we have proposed a wave propagation model to take into account effects of operational conditions such as internal stresses caused by mechanical loading. In this work, we develop a strategy for estimating these load-induced (large) deformations from ultrasonic measurements. In the context of least-squares optimization, we minimize the difference between measurements and the observed direct model. Among the available methods there exist variational methods, such as Full Waveform Inversion , and sequential approaches, such as Kalman Filtering. As the linearized aspect of the direct model leads to an unwieldy tangent dynamics, tangent-free methods are preferable. After reducing the dimension of the parametric space by decomposing the deformation on selected eigenmodes of a static problem, we apply iteratively the Reduced-Order Unscented Kalman Filter as estimation method. We will show how this method can be understood as a gradient free alternative to an iterated Gauss-Newton based minimisation of the least squares functional, or incremental 4D-variational method. Moreover it allows to launch multiple direct solvers in parallel leading to an efficient exploration of the parametric space and sensitivity analysis w.r.t. extended sets of eigenmodes. We illustrate the proposed inversion strategy by applying it to realistic cases.