Damage assessment of connection in a framed structure
using a system identification algorithm
Kyu Chul Lee
This paper presents a damage assessment technique for the connection of framed structure using a system identification. Because connections are more susceptible to damage than the other parts of structure in a steel structure, not element but connection damage model is used, which is more realistic. The concept of semi-rigid connection is utilized to represent the reduction of connection's rotational stiffness due to damage. System identification using static responses of structure has discontinuity and non-uniqueness of solution caused by measurement error and sparseness of measurement data, respectively, which make the solution unstable. Particularly, structure is so large that it is difficult to avoid the sparseness. This paper focuses on rank-deficiency. Tikhonov regularization technique and TSVD have been widely used to overcome instabilities of system identification. Generalized Tikhonov regularization is proposed to solve difficulties in using TSVD. L1-regularization is also suggested to overcome the limit of L2-regularization. Statistical damage assessment is used for more reliable estimation. Numerical examples verify the validity of this paper.
system identification, semi-rigid connection, rank-deficiency, Tikhonov regularization, TSVD, generalized Tikhonov regularization, L1-regularization, statistical damage assessment