Data-driven method for real-time reconstruction of antenna element displacement

Real-time reconstruction of antenna element displacements using limited strain data is essential for mitigating the degradation of phased array radar systems' electrical performance caused by antenna deformations. This study presents a data-driven framework for efficiently and accurately reconstructing antenna element displacements. The proposed approach combines finite element simulations with a multilayer perceptron (MLP) network to predict the displacements of all target points based on strain values from a limited number of monitoring points. The MLP network's architecture is optimized using a loss function designed under the assumptions of Gaussian distribution and homoscedastic uncertainty. Displacement reconstruction for several target locations on a cantilever plate and a four-sided hinged support plate is demonstrated, utilizing a database generated through finite element analysis (FEA). Simulated discrete strain data derived from FEA are employed as input to the MLP-based shape sensing model (SS-MLP), effectively replicating real-world strain measurements. A comparative analysis between the SS-MLP and FEA models reveals that this method achieves precise and simultaneous displacement reconstruction for multiple points, even with sparsely distributed monitoring locations. These results suggest that the proposed approach provides a promising alternative for real-time antenna element displacement monitoring.