Plate and shell components are important elements of safety-critical structures, such as airplane fuselages or wind turbine structures. Structural Health Monitoring (SHM) schemes can be used to track the condition of these systems and warn of impending damage. To this end, Guided Wave (GW) based methods, relying on Lamb waves, feature appealing traits, such as low geometrical damping, which allows to cover a large monitoring area, with few sensors. A further advantage stems from their short wavelength, which enables the detection of even small damages. The short wavelength, however, requires a fine discretization in numerical solvers, which leads to a high computational effort in modeling propagation of Lamb waves. This poses a hurdle towards the application of hybrid modeling schemes, which require fusion of measurement data with numerical models for succeeding in damage identification and localisation tasks. In this work, we propose a Reduced Order Model (ROM), able to simulate the wave propagation at reduced computational toll, by employing a parameterization with respect to possible damage locations in the frequency domain. The adopted principle exploits the interaction of a wave with a defect, as this generates new waves due to reflection and mode conversion and can be viewed as a new source of wave generation. The ROM construction exploits the principle of Frequency Response Functions (FRFs). We demonstrate use of this FRF-based ROM in an inverse problem via a synthetic example of a plate featuring different types of defects, such as notches or conical holes.