Total least-squares criteria in parameter identification for flight flutter testing

Abstract

An investigation of parameter identification of time series models for linear dynamic structural systems using the least squares and total least-squares criteria is presented. Excitation and response time-domain data are used to parameterize autoregressive, moving average models. The method, or criteria, which is used to solve a set of overdetermine d, linear algebraic equations developed from the time-domain data affects the solution. A commonly used criteria, least squares, introduces the possibility of significant bias error hi the system parameters and leads to bias errors in the modal parameter estimates when measurement errors are present. An alternative criteria, total least squares, provides an approach that appears to significantly reduce the bias error in the parameter estimates. These methods are applied to a simple, simulated system and then to flight flutter test data with particular emphasis on accurate modal damping estimates. [A] [A'] [4'.>'