High-dimensional systems often exhibit low-dimensional structure that can be exploited for efficient state estimation. When full measurements are impractical, we seek a small number of well-chosen sensors that allow accurate reconstruction of the system state. In here, we present the usage of the data-driven optimal sensor placement (OSP) framework of Manohar et al. (2018), to predict the sensor locations on structural problems based on the full-order FE dynamic solution as the input snap-shots. The methodology uses singular value decomposition (SVD) of the snap-shots to construct the modes containing most of the energy in the full-order solution, followed by QR pivoting to compute the optimal sensor placements. An extension for the case where some sensors are preselected and additional sensors must be optimized is also presented. The OSP procedure is demonstrated on a simply supported Kirchhoff-Love plate problem, where the full-order solution is obtained with the isogeometric finite element solver IFEM. The procedure is also demonstrated on a much more complex shell finite element model of a ship (the Gunnerus vessel).