Comparison Of Structural Damage Detection Algorithms Based On The ASCE SHM Benchmark Structure

Research of Structure Health Monitoring (SHM) is one of the top issues in civil engineering, and at the same time, structural damage detection is a core problem of SHM. A large number of methods for structural damage detection are proposed so far, and each method has its own advantages and disadvantages. Facing the damage detection problem in real situation, selection of appropriate methods becomes a concerned problem. In this paper, several types of damage detection methods are tested against the ASCE SHM Benchmark model, and they are evaluated accordingly. These results will provide insight for practitioners to select appropriate algorithms to deal with damage detection problems.Firstly, the SHM benchmark model which was proposed by American Society of Civil Engineers (ASCE) is introduced. Three types of damage detection methods, i.e., Flexibility difference method, Sensitivity analysis method and Element mode strain energy method are employed to detect the damages in the benchmark model with simulated results. The feature and suitability are compared accordingly.Secondly, Artificial Neural Network approach is employed for damage identification of the benchmark structure. Normalized frequency change rate (NFCR) and frequency change rate (FCR) are used as input vectors to artificial neural network. Then structure and learning algorism of artificial neural network for damage detection are determined and appropriate network parameters are selected. Using BP network with momentum item, damage identification of the benchmark structure is implemented in three steps, which include locating damaged layer, identifying damaged member in detected layer and identifying the damaged member's severity.Lastly, by analyzing the damage identification results for the benchmark model, the applicability of method of each specific damage level is provided by comparing their merits and demerits.