Modeling and simulating guided wave propagation in complex, geometric structures is a topic of significant interest in structural health monitoring. These models have the potential to benefit damage detection, localization, and characterization in structures where traditional algorithms fail. Numerical modelling (for example, using finite element or semi-analytical finite element methods) is a popular approach for simulating complex wave behavior. Yet, using these models to improve experimental data analysis remains difficult. Numerical simulations and experimental data rarely match due to uncertainty in the properties of the structures and the guided waves traveling within them. As a result, there is a significant need to reduce this uncertainty by incorporating experimental data into the models. In this paper, we present a dictionary learning framework to address this challenge. Specifically, use dictionary learning to combine numerical wavefield simulations with 24 simulated guided wave measurements with different frequency-dependent velocity characteristics (emulating an experimental system) to make accurate, global predictions about experimental wave behavior. From just 24 measurements, we show that we can predict and extrapolate guided wave behavior with accuracies greater than 92%.