Designers use simulations to observe the behaviour of a system and to make design decisions in order to improve dynamic performance. However, for complex dynamic systems, these simulations are often time-consuming. For robust design purposes, numerous simulations are required as the range of design variables are investigated in search of the optimum set. This paper investigates the use of both singular value decomposition and metamodeling to reduce the time required to calculate the time history of the response. Singular value decomposition splits the time-history information into both parameter space and time space so that metamodels in the form of polynomial interpolation functions are fit for only a few columns of the parameter space. Both Regression and Kriging models are investigated and compared for their accuracy and time required for preparation. Importance analysis through sensitivity functions further helps reduce time. This overall reduction in time makes it feasible to find a robust design through maximizing probability of conformance at selected times. The practicality and potential of the approach is demonstrated by parameter design of a linear servo-system.