A reduced order model based on singular value decomposition and correlation is developed to capture the nonlinear dynamics of a hypersonic vehicle in flight. A set of training samples in state space form are collected using the complex step method. These samples are used to identify a set of ordered bases which describe the variation of the state matrices and state rates. Surrogate functions are fit to the coefficients of these bases to approximate the training samples and predict the state matrices outside of the training set. The dynamics of the system may then be rapidly simulated, provided the training set sufficiently populates the state space. This approach is first applied to a spring-mass-damper system with variable nonlinearity and number of degrees of freedom to determine training sample size and surrogate function orders which produce stable and accurate time simulation. A representative hypersonic vehicle is then considered to demonstrate the potential of this approach for rapid simulation of hypersonic flight.
展开▼
