Efficient reduced-order system identification for linear systems with multiple inputs

A new, efficient discrete-time-domain system identification and model reduction method for a large-scaled linear dynamic system with multiple inputs is presented. The method is based on a modification of the classical eigensystem realization algorithm and a simultaneous injection of multiple inputs, so called the single-composite-input method. Because the system response is sampled almost exclusively for the single representative input, this technique can significantly reduce the model construction time as well as the amount of the sampled data. For derivation of the new algorithm, the singular value decomposition is performed using output measurements that are not necessarily attributed to pulse inputs. Application to general computational-fluid-dynamic systems and formulation of reduced-order aeroelastic models are also presented. The efficiency and accuracy of the method is demonstrated via a simple aeroelastic system, which is modeled by the three-dimensional vortex lattice and a six-degrees-of-freedom plate-like structure.