Influence of incomplete modal data on the performance of a vibration-based damage detection methodology

The use of changes in the dynamic response of a structure after damage has become very popular in the scientific community to formulate methodologies that permit assessing the integrity of a structure. In this sense, a finite-element model (FEM) that represents the damaged structure can be obtained by minimizing the difference between the dynamic response of this model and the response of the current structure obtained experimentally. The optimization variables are composed of the stiffness reduction factor of each element that belongs to the FEM of the undamaged structure. This paper proposes the use of an adaptive Particle Swarm Optimizer to solve the optimization problem associated with the detection of the damage in a multi-supported beam structure and studies how incomplete data affect the performance of the proposed algorithm. Adaptation is implemented to avoid the definition of the PSO parameters - cognitive and social parameters - by trial and error. Natural frequencies and mode shapes were selected as the dynamic characteristics to be used in the objective function. As this research was developed by using numerical simulations, information about only a few first modes in specific degrees of freedom was considered available in order to take into account the incomplete measurement issue. Results have shown that a minimum quantity of modal data is necessary to guarantee the success of the damage detection methodology and that the ability to locate and quantify damage may not be improved by using excessive information. It has also been observed than simple damage scenarios can be more reliably detected than multiple ones.