Model updating‐based automated damage detection of concrete‐encased composite column‐beam connections

Summary This paper presents a comparative study about the numerical and experimental system identification and model updating‐based automated damage detection of concrete‐encased composite column‐beam connections. Four different column‐beam connection types named as CBC#A, CBC#B, CBC#C, and CBC#D were considered without any changes in geometrical configuration. Numerical and experimental dynamic characteristics were extracted by initial finite element analysis and nondestructive experimental measurements for undamaged condition. Enhanced frequency domain decomposition (EFDD) method was used for experimental modal extractions. A good agreement was observed between experimental natural frequencies but not enough correlation between damping ratios for all connection types. Maximum differences between initial finite element and undamaged experimental results were calculated as 34.68% for CBC#A, 40.78% for CBC#B, 33.15% for CBC#C, and 47.13% for CBC#D type connection details. The differences were reduced to 0.24%, 0.24%, 0.16%, and 0.15%, respectively, by automated model updating procedure using spring coefficient, modulus of elasticity and weight per unit volume. In the case of the damaged condition, lateral forces were applied to simulate the effect of an earthquake. Cracks strongly affected the natural frequencies. The mode shapes were not broken evidently after damaged conditions. The mode shapes definitely deviated a bit, but generally the same after damages. The natural frequencies have decreased non‐monotonically due to the decrease in the rigidity of the column and beam at the cracked section. Maximum differences were calculated as 14.96% for CBC#A, 41.08% for CBC#B, 31.35% for CBC#C, and 45.93% for CBC#D type connection details. After automated model updating the above‐mentioned differences were reduced to 0.35%, 0.78%, 0.73%, and 0.85% respectively. Contour diagrams of changes of updating parameters were plotted for damage identification. The regions with high difference rates indicated the location of the damage.