Time-domain identification of linear dynamic structural systems using total least squares

The use of time domain methods to estimate modal characteristics of linear, dynamic structural systems is investigated. The natural frequency and damping of the fundamental vibration modes are the modal parameters which are of primary interest. The research is motivated by the need to assess the aeroelastic stability of aircraft during flight flutter tests. To be effective in the flight flutter test environment, an identification method must use short data records to accurately estimate the damping of closely spaced vibration modes in a limited amount of time.;The complete identification process includes the acquisition of excitation and response data from a modal test, the selection of a model for the system, the identification of the model parameters using the experimental data, and the estimation of modal characteristics from the identified model. Frequency domain techniques have traditionally been used for these parameter identification problems but time domain techniques are considered in this research effort as an alternative approach.;The time domain methods investigated in this dissertation are based on linear difference equations and are identified with discrete time domain data. Although previous efforts have used these models successfully to estimate modal parameters, persistent bias errors in the damping estimates have restricted practical implementation of the methods. A significant portion of this dissertation is dedicated toward identifying the sources of both bias and precision error and developing techniques to reduce the errors. Averaging of estimates and signal enhancement techniques are shown to be successful methods of reducing the precision error. The total least squares criteria and order overspecification are methods of significantly reducing the bias errors. An improved time domain identification method, backward total least squares, has been developed as an extension to previous identification methods. Its effectiveness is evaluated with data from numerically simulated dynamic systems and with experimental data from flight flutter tests.