We examine the question of uniqueness of solution to the modal paramet
er problem by applying a least-squares output error estimator to a six
-degree-of-freedom shear building using both a two- and a six-paramete
r model. Multiple solutions are located by initiating the iterative pa
rameter estimation algorithm from random starting points. The case stu
dy provides tangible illustrations of the phenomena latent in modal pa
rameter estimators and demonstrates several counter-intuitive results.
The study of the two-parameter model allows us to show the connection
between multiple solutions for noise-free data and solution clusterin
g for noisy data. The results suggest a new definition of the best mea
surement set as being the one most likely to lead to the correct solut
ion with the least sensitivity to noise. We conclude that modal parame
ter estimation from sparse measurements, albeit fraught with difficult
ies, can yield valuable results. (C) 1996 Academic Press Limited