Impulse response functions are commonly used as the raw data for time domain modal parameter identification. It is well known that the noise-signal ratio of such time signals gets progressively worse as the decay progresses. Most of these identification algorithms have the Least Squares method as their fundamental component, and consequently model order overspecification has to be used in order reduce the bias on the estimates. This paper explores the use of exponential weighting in the solution to reduce the amount of overspecification needed to obtain accurate modal parameter estimates. The approach can be used on a wide variety of different time domain algorithms, and is illustrated using simulated data sets.
展开▼
