Structural Damage Identification

In recent years, structural health monitoring has become a hot topic of research for civil infrastructures, buildings, mechanical systems, aircraft and aerospace structures. In particular, it would be very useful if some scientific and systematic methods can be developed to detect and even quantify damage in structures. Damage may be due to deterioration over time, or caused by natural disasters or man-made actions. If not detected early, such damage would increase maintenance cost, disrupt operations and even cause catastrophic failure. It is therefore essential to carry out regular monitoring for early detection and rectification of structural damage. 

Funded by the NUS, a research project was carried out to formulate a numerical strategy for structure assessment in a quantifiable and objective way. Involving both numerical and experimental studies, the emphasis is on non-destructive damage detection of structural members. A time-domain system identification method, namely the extended Kalman Filter method, is adopted with several significant modifications. To reduce the identification system non-linearity, the method of stiffness-damping uncoupling is introduced which gives significant improvement for damage identification with incomplete and noisy measurements. The well-known “leakage” problem is reduced by repeating the identification procedure on different windows of I/O time signals and taking the average results. Two dimensionless indices are used to reflect the extent of damage, in addition to detection of damage location.

Figure 1: Identified damage in terms of stiffness correction factor (SCF) 
under 10% I/O noise (1-6: beam elements; 7-8: end rotational springs).
 


Figure 2: Experimental set-up for verification of the proposed damage identification strategy.

In the numerical study, a structural member is divided into six beam elements with two end rotational springs. The rotational springs serve to simulate the unknown semi-rigidity at the two ends of the structural member. Thus, the structural parameters to be identified include the flexural rigidity values of all elements, two damping parameters and two the end rotational spring constants. For example, local damage of 50% and 30% are assumed to take place at elements 2 and 5, respectively. Figures 1 shows the identification results under 10% input and output noise. The damage assessment results as reflected by the dimensionless stiffness correction factor (SCF), give a distinct indication of the damage locations and severity. Quantification of damage extent is satisfactory and leakage to undamaged element is acceptably small. 

The numerical study is complemented by an experimental study, involving vibration testing for pre-damage and post-damage states. Most research work on structural damage identification has dealt with only numerical studies and have not been subjected to real tests. Experimental investigation serves as a more severe test than the numerical study. An experimental study involving shaker-excited vibration tests of a reinforced concrete beam was performed in the laboratory (Figure 2). The experimentally identified change in the flexural stiffness (EI) value of the third element is found to be 32% and is reasonably close to the estimate value of 40%. 

In conclusion, both numerical and experimental studies have shown that the SCF values obtained by the proposed damage identification strategy gives a clear indication of the location and severity of local damage.