Kevin Truman and Gus Terlaje

Department of Civil Engineering, Washington University in St. Louis, USA
E-mail: at1@cec. wustl. edu


Damage of a structure is caused by many factors, some of which include earthquakes, wind, snow, increased live loads and simply the age of a structure in the form of fatigue or deterioration of its structural elements. Early detection of damage is vital in order to prevent the continuous degradation of the structure, ultimately leading to a catastrophic failure of the system. Traditional methods of detecting this damage are invasive and often require partial destruction of non-structural elements, such as wall panels and finish work. This detection process is executed blindly with little a priori knowledge of the possible location or extent of damage. Furthermore, damage often is not detected until it becomes so extensive that it is visible by building occupants. Structural health monitoring is employed to avoid unnecessary demolition through accurate prediction of damage location.


Even though the area of structural health monitoring has been studied extensively in recent years, the majority of research and implementation has been based on damage detection using dynamic properties of the system. Using changes in vibration characteristics of a structure, damage may be detected and located (Salawu 1997). These vibrational characteristics are dependent on several system parameters such as; mass, stiffness, and damping, all of which require time and effort to obtain. Dynamic structural health monitoring has yielded significant results and therefore extensive research is justified, however requiring multiple system parameters creates inherent challenges in dynamic analysis. Because of the difficulties associated with dynamic analysis, a process using static properties merits exploration. Several methods have been developed using static responses to detect damage. Some methods combine both static and dynamic responses in order to locate damage. One such method predicts the location of damage by attempting to correlate expected and actual damage signs using first-order approximations for changes in static displacements and natural frequencies. Next a separate routine is used to determine the extent of damage at the predicted location (Wang et al.,2001). Although the results of such a procedure have been shown to be accurate for a planar truss and fixed-fixed beam, the effectiveness of the procedure for more complicated structures is not well known. Still other methods have been explored using solely static data. Such methods have investigated the correlation between measured displacements at various degrees of freedom and applied forces at other degrees of freedom in order to assemble the stiffness matrix of the system (Sanayei et al. 1991). Approximation techniques can then be used to obtain unknown displacement values. Such methods have accurately located and quantified damage for 2-D truss and beam element frames. A similar technique has been developed using unconstrained nonlinear optimization (Johnson et al. 2004). Displacements are incorporated into an error function which is dependent on only the cross sectional properties of each structural member. This method successfully detected damage in continuous beams and multi-bay, multi-story frames. Again, the effectiveness of this method is not known for other structural systems. The current theory will also present a method of detecting and quantifying damage using only static measurements in order to assemble the stiffness matrix of the structure. Once the damaged stiffness matrix is known, comparison with the healthy stiffness matrix can locate and quantify damage using the cross sectional properties of each structural element.


M. Pandey et al. (eds), Advances in Engineering Structures, Mechanics & Construction, 685-696. © 2006 Springer. Printed in the Netherlands.