Machine learning and artificial intelligence has been applied to other facets of structural mechanics and structural dynamics, mainly for structural optimization and structural reliability analysis, but have seen little use in surrogate modeling for structural time series prediction. The current research will focus on how data reduction tools (such as mathematical morphology) can be applied to dynamic structural data and can be used in conjunction with clustering methodologies as well as artificial neural networks (ANNs) to create a useful and highly computationally efficient dynamic model that can be used to make predictions on the acceleration response of a structural system. The current study utilizes training data developed from finite element modeling of a simple system that demonstrates a nonlinear behavior of interest (a change in the dynamic response with a change in loading magnitude), but the methodologies developed could be applied to other modeling or analysis schemes, test data, or in-situ measurements of a structure of interest. The results of the study show how mathematical morphology tools can reduce the dimensionality of time series data while still preserving important characteristics like natural frequency, and the use of ANNs shows promise as a surrogate model for dynamic response prediction.