Structural Damage Detection Based On The Identification Of Physical Parameters

The structural damage detection has been a hot research topic in the field of civil engineering. In the recent decades, researchers have been exploring a number of damage detection methods that are feasible for whole structure. Their basic ideas are (1) determine the structural mechanics parameters by structural dynamic response; (2) detect structure damage according to the variation of the structural mechanics parameters. The feasibility of various mechanical parameters for structure damage detection has been analyzed, which showed that the structural stiffness is the best suitable physical parameters for damage detection, and the corresponding theory and method of physical parameters identification have been studied in detail according to this idea.Kalman Filter, Extended Kalman Filter and Particle Filter, discussed in this paper, are widely used for the identification of physical parameters of structure. The theoretical analysis and numerical simulation show that there are some shortcomings among the three filter techniques: (1) they can only be used for lumped mass shear-type structure, (2) their identification results are not accurate enough and are strongly influenced by the settings of initial parameters, which is difficult to derive regularity from, (3) The deviation of a few results is relative high during the test without noise, while the deviation of most of results are very high during the test with noise. . The three filter techniques were not chosen as the main study of this paper because of their limit restrict of application.The methods of detecting structural physical parameters using single-objective intelligent optimization algorithms are well developed during the recent decades and become a hot research topic. Four single-objective functions defined by natural frequencies, modes, flexibility matrix and acceleration response were analyzed, and Genetic Algorithm (GA), Particle Swarm Optimization Algorithm (PSO) and Differential Evolution Algorithm (DE) were introduced. Numerical simulations show that the damage detection method consisting of DE and the single objective function defined by acceleration response has high precision and good anti-noise ability, but its disadvantages are the uncertain of initial searching ranges, low computation efficiency and lack of online identification capability. In order to solve these disadvantages, an equivalent single degree method was proposed and verified by numerical simulation.Because of the complexity and diversity of many optimization problems, multi-objective optimization algorithms are more suitable to solve these problems than single-objective optimization algorithms. Nevertheless, relatively few studies have been reported on structural identification, and many multi-objective intelligent optimization algorithms appear to be popular choices. Three multi-objective functions defined by natural frequencies, modes, Modal Assurance Criterion (MAC) and acceleration response were proposed, and multi-objective GA (NSGAII), multi-objective PSO (CMOPSO) and multi-objective DE were introduced. Numerical simulations showed that the damage detection method consisting of DEMO and multi-objective function defined by natural frequencies and summed MAC has high precision and computation efficiency, but its anti-noise ability is poor; the damage detection method consisting of DEMO and multi-objective function defined by natural frequencies and acceleration response has high precision and good anti-noise ability, but its computation efficiency is low. These two damage detection methods are standout, although they have disadvantages.In order to verify two multi-objective functions and the equivalent single degree method proposed in the paper, a truss numerical model consisting of 31 elements and 1:1 Reinforced Concrete (RC) structure used by shaking-table test were analyzed. The analyzed results showed that these multi-objective functions have higher precision and better stability than multi objective function proposed by other author, and although the equivalent single degree method leaves something to be desired, it is practical meaning in damage detection.