PSI - Issue 37

Elizabeth K. Ervin et al. / Procedia Structural Integrity 37 (2022) 6–16 Ervin and Zeng / Structural Integrity Procedia 00 (2021) 000 – 000

7

2

should be conducted periodically to ensure structural integrity, and agencies set regulations to do exactly that. More efficie nt detection techniques could assist inspectors detecting dangerous levels of global damage so that collapses and casualties would be less likely. Underfunding and poor allocation can prevent comprehensive inspections, especially for lower value structures. Global measurement of civil engineering structures can be conducted via vibration-based techniques. Such techniques generally determine changes in physical properties of a structure, i.e. mass, stiffness, or damping, by tracking the variations in modal properties [Talebinejad et al. (2011), Oskoui et al. (2019)]. Damage sensitive features (DSFs) are certain quantities extracted from structural response that are sensitive to the presence of damage [Farrar and Worden (2012)]. These DSFs can be utilized to generate secondary features that are expected to better correlate with damage than direct sensor magnitudes. A common approach to detect damage is comparing two states of the same structure. The two response data sets are used to generate DSFs of the two states, and change of the same DSF is a damage index (DI), a mathematical method to quantify damage location and extent. Although several DIs have been proposed in literature, the accuracy and consistency of a single DI could be affected by various of factors, such as damage extent, damage types, structural complexity and so on. Therefore, using multiple DIs for damage detection is expected to be more robust and accurate. This work proposes a new damage identification approach by combining various DIs (including new ones and all three directions) into one single optimized DI. To this end, an artificial intelligence computational technique is required. GAs have also been applied to structural damage identification. Rao et al. (2004) proposed a GA-based approach for damage localization and quantification within structural components using the concept of residual force. The GA procedure is verified by applications including a plane truss, a cantilever beam, and a portal truss. Chou and Ghaboussi (2001) developed a GA-based method for detection and identification of damage in truss bridges. Section properties of truss members are encoded as the GA’s genes, and the fitness function is formulated by the difference of measured and computed static displacements. Damage thus could be identified by the section properties that minimize the fitness function. Perera and Torres (2006) proposed a global damage detection and assessment methodology for damage location and severity detection using a GA technique. The method is applied to a simply supported beam and verified by both numerical simulation and experimental data. None of these studies considered tri-axial data or multi-dimensional DIs. This study introduces a robust GA-based damage quantification methodology to detect both location and severity using modal information from common civil engineering structures. Nomenclature d, * Damaged P Probability ̈ Measured acceleration response P Weighted average combining DIs U Undamaged R Resultant COMAC Coordinate modal assurance criterion S Spatial CDF Cumulative density function T Target Vector DI Damage index named by (method)(DSF)(R or S) U,  Strain Energy Diff , Div , Perc Method applied to DSF for DI Z Z-score DL Damage Level  Weight coefficient, DI participation DSF Damage sensitive feature  Absolute difference between mean values EI Young’s modulus and moment of inertia  Curvature F Flexibility  Mode shape GA Genetic algorithm  Maximum reduction of E FV Fitness value  Mean value  Natural frequency 2. Procedure of damage localization This section presents the procedure of utilizing time-history data or finite element (FE) model output to generate twenty-four metrics and one plot that provide in-depth information about damage condition to inspectors. Multiple indices, from literature review or newly proposed in this paper, are developed on the basis of features to quantify