A physics-based virtual sensing framework integrating Kalman filter and Newmark-beta method for structural dynamics

Abstract

Accurate structural dynamics analysis depends on the reliability of numerical models, which are commonly validated through modal testing. In practice, however, sensor placement during testing is often limited by spatial and environmental constraints, leading to incomplete measurement data. To address these limitations, virtual sensing has emerged as an effective solution for estimating structural responses at unmeasured locations. This study proposes a physicsbased virtual sensing framework that integrates Kalman filter algorithm with a Newmarkbeta implicit time integration to reconstruct full-field response. A modal reducedorder model is employed to improve computational efficiency. Unlike virtual sensing approaches that focus on inverse force identification, this framework reconstructs unmeasured acceleration responses under measured excitation, which is more relevant to test and validation processes. The proposed method enables accurate identification of dominant natural frequencies using limited sensor data, without requiring extensive test-simulation correlation. The framework is validated through shaker tests on an automotive chassis crossmember, where the reconstructed responses show agreement with experimental results in the frequency domain. The proposed framework provides a practical solution for reliable response estimation while reducing the number of physical sensors and correlation processes required in structural dynamic testing.