Contribution
Computationally Efficient Virtual Sensing Method for Exterior Vibroacoustics Using a Time-Domain Finite Element Model with Perfectly Matched Layers
* Presenting author
Abstract:
Virtual sensing is an online estimation framework that can minimize physical sensor deployment and estimate full-field responses based on a reliable simulation model. Building the simulation model for exterior vibroacoustic problems in unbounded domains presents significant computational challenges due to the need for extensive meshing and accurate modelling of infinite domain conditions. Previous work from the authors developed a time-stable vibroacoustic finite element (FE) formulation using the locally conformal perfectly matched layer (LCPML) method. The LCPML enables the use of general convex enclosures around the vibroacoustic FE model, effectively absorbing outgoing waves to simulate the infinite boundary condition. Additionally, an automatic Krylov-subspace model order reduction (MOR) technique is applied to the model, further reducing computational effort. Based on this efficient and time-stable vibroacoustic FE model, a virtual sensing framework is developed by using a Kalman filter to estimate the radiated sound pressure and intensity field from limited vibration measurements. This framework was numerically tested on a free radiation model from a plate structure using simulated measurement data with artificial noise, to demonstrate its ability in computational savings and real-time, full-field estimation of vibroacoustic responses.