Proceedings of 8 th International Conference on Short and Medium Span Bridges Niagara Falls, Canada 2010 143-1 DAMAGE DETECTION FOR TRUSS BRIDGE STRUCTURES USING CORRELATION-BASED STRUCTURAL MODAL STRAIN ENERGY Liang Wang School of Urban Development, Queensland University of Technology, Australia Tommy H.T. Chan School of Urban Development, Queensland University of Technology, Australia David P. Thambiratnam School of Urban Development, Queensland University of Technology, Australia Andy C.C. Tan School of Engineering System, Queensland University of Technology, Australia ABSTRACT This paper presents the feasibility of using structural modal strain energy as a parameter employed in correlation- based damage detection method for truss bridge structures. It is an extension of the damage detection adopting Multiple Damage Location Assurance Criterion (MDLAC). In this paper, the sensitivity of modal strain energy to damage, from the analytical model, is incorporated into the MDLAC method. Firstly, the sensitivity matrix of modal strain energy to damage is conducted offline. For an arbitrary damage case, the fitness function (MDLAC) is calculated by multiplying the sensitivity matrix and damage vector. Then genetic algorithm (GA) is used to iteratively search the damage vector that maximises the correlation between the corresponding modal strain energy change (hypothesised) and its counterpart in measurement. The proposed method is simulated and compared with the conventional methods, e.g. frequency-error method, coordinate modal assurance criterion and MDLAC using mode shapes on a numerical truss bridge structure. The results demonstrate the MDLAC using modal strain energy method can yield acceptable damage detection outcomes with less computing efforts, even in a noise contaminated condition. 1. INTRODUCTION Vibration-based damage detection (VBDD) has received considerable attention in recent decades. It stands on the basis that damages in physical models will lead to structural stiffness decrease and consequently changes in vibration parameters, e.g. natural frequencies, mode shapes and their derivatives. Identifying these changes could be helpful to determine a structure’s damage condition. Major techniques used in VBDD make use of modal parameters identification according to the finite element (FE) model. Some literature has comprehensively reviewed the damage detection approaches based on structural vibration characteristics (Doebling et al., 1996a; Sohn et al., 2004; Wang et al., 2009). For example, natural frequency has been widely known as one of the parameters with less contamination from measurement noise and could be easily measured. Cawley and Adams (1979) adopt a damage indicator using the frequency changes of two modes though it is proved to be location-sensitive only. M. Biswas et al. (1990) develop the experiments for a highway bridge and find that even using natural frequency change alone could detect damage. However, in practice, only a limited number of modal frequencies are obtainable and the changes of them before and after damage are insufficient to reveal damage information of local members, because the different combinations of damaged members might result in the same frequency change. Mode shapes and its derivatives contain the spatial dynamic