Department of Civil Engineering,Bengal Engineering and Science University
Abstract
A sensitivity based parameter identi cation method is presented to detect the damage of existing
structures, using applied sets of static forces at one subset of degrees of freedom and measured
displacements at a subset of degrees of freedom that may overlap completely, partially, or not
at all. The algorithm follows an output error approach, which minimizes the deviation between
a measured and a theoretical displacement, in lieu of the commonly used force error function.
An iterative scheme is developed utilizing a rst order Taylor series expansion to linearize the
associated non linear problem. The algorithm automatically adjusts the structural element
sti ness parameters, in order to improve the comparison between a measured and a theoretical
response in an optimal way. The measured input required in the present study is arti cially
generated. The e ect that a noisy displacement measurement has on an identi cation procedure
is also studied. A procedure is also identi ed, in order to select the limited number of DOF
required to perform successful parameter identi cation, reducing the impact of measurement
errors on the identi ed parameters. The algorithm is elucidated by a numerical example on
frame structures.
)