With a sufficiently great number of sensors and actuators, any finite dimensional dynamic system is identifiable on the basis of input-output data. It is presently indicated that, for conservative nongyroscopic linear mechanical systems, the number of sensors and actuators required for identifiability is very large, where 'identifiability' is understood as a unique determination of the mass and stiffness matrices. The required number of sensors and actuators drops by a factor of two, given a relaxation of the identifiability criterion so that identification can fail only if the system parameters being identified lie in a set of measure zero. When the mass matrix is known a priori, this additional information does not significantly affect the requirements for guaranteed identifiability, though the number of parameters to be determined is reduced by a factor of two.
Identifiability of conservative linear mechanical systems
1985-03-01
Miscellaneous
No indication
English
Parameter identifiability of linear dynamical systems
NTRS | 1974
|Near Identifiability of Dynamical Systems
NTRS | 1987
|Identifiability Of Systems With Modeling Errors
NTRS | 1988
|British Library Conference Proceedings | 1995
|Nonlinear identifiability: an algebraic approach
IEEE | 2003
|