Dynamic strain response measurement‐based damage identification in structural frames

It has been a common practice to utilize measured linear acceleration responses for structural health monitoring primarily because it is perhaps easiest to collect data on these. This paper, on the other hand, is focused on the utilization of measured strain responses as they provided more complete and useful information for health monitoring, especially for structural frames. The measured strain responses are used to extract strain mode shapes and frequencies of the system using the well-known Stochastic System Identification or NExT-ERA algorithms. The extracted strain mode shapes and frequencies are then utilized to identify the element stiffness characteristics using a two-step pseudoinversion approach. The first pseudoinversion step involves the calculation of the least squares estimates of the system mode shapes from the strain mode shapes followed by the second step of calculating the element stiffness values again by second pseudoinversion using single or multiple system modes. Numerical results are presented to examine the effects of the two inversion steps, the utilization of single and multiple modes, and the measurement error on the accuracy of the calculated stiffness values and on the identification of the damaged elements. It is observed that in framed structures with asymmetric configuration of structural elements, it may be necessary to use multiple modes with optimally selected mode contributions factors to accurately identify the damage levels and locations. The proposed approach is seen to work very well in identifying the locations and levels of damage in multiple structural elements in a frame.