Theoretical And Experimental Studies Of Structural Damage Identification Based On Energy Indexes

With the social development, scientific progress and aggrandizement of people's safety requirement, structural damage identification is attracting more attention and becomes a new research focus in the field of civil engineering, because it does play an important role in the operation of protecting people's life and property. Therefore, it is a significant work to study on this subject.Many damage identification methods have been developed and, most of them are frequency-domain ones which possess some limitations in practical applications. In contrast, time-domain methods utilize the measured signals to detect structural damage in time-domain directly, requiring no perfect structural modal parameters and preserving the original time-domain information of tested signals. As a result, they are suitable for the real-time and online structural health monitoring. At the same time, energy indexes are sensitive to structural damages which induce the structural local energy changes. Thereby, energy indexes-based methods do well in this subject and they do not need structural model building and modal shape analyzing. Based on aforementioned background and supported by the National Natural Science Foundation of China (No. 50378041 & 50778077), in this dissertation, theoretical and experimental studies of structural damage identification based on energy indexes have been implemented.In this dissertation, following aspects have been studied theoretically, numerically and experimentally and, some important results and conclusions have been acquired:(1) The statistical moment, which has been used to develop a new damage detection method for shear building structures, is a time-domain energy index essentially. It is a good damage index with strong noise robustness. However, it is only effective for shear type structures but ineffective for bending type ones. A progress is made to the index for damage identification in bending type structures here using a two-step method: first, the locations of damaged elements are detected utilizing a new index proposed here, named statistical moment curvature; then, the damage extent of these elements is identified with a model updating method. The effectiveness of this new method is verified with a numerical example.(2) A new damage detection technique based on kinetic energy density, which is firstly developed applying the tested acceleration signals, is presented here. Associating with the rotational spring model, it identifies both the crack location and depth in a beam model successfully. Then, the limitation of this method is pointed out and, overcomed with some techniques such as wavelet packet transform and correction of datum line. The diagnosis results of a numerical continuous beam and a 3D steel frame confirm the damage detection ability and noise robustness of this new method.(3) A theoretical analysis based on a cantilever is accomplished to prove the following argues: structural damages cause strain changes in the corresponding locations, which implies a one-to-one correspondence between the strain changes and damage locations; however, displacement changes do not possess this property because a damage occurring in one element could induce displacement variation of another element. Thus, not the displacement type parameters but the strain ones are good for damage localization.(4) A strategy using strain test data for damage detection is proposed by defining a new index named "Pseudo Strain Energy Density (PSED)". Applying the gapped-smoothing method, the undamaged PSED curve can be obtained from the damaged one and this new strategy becomes a non-baseline one which is significant to the damage estimation when lacking the undamaged information. Furthermore, two other indexes, the statistical PSED and PSED grey relation coefficient, are addressed for damage localization. By combining the PSED method with the genetic algorithm, both the damage location and damage extent can be identified and the analysis time decreases sharply.(5) Wavelet packet transform is proved to be a strong tool for signal preprocessing. In these damage identification methods, it is utilized to accomplish signal decomposition and recomposition. During this process, the most useful signal components are extracted and used for damage detection. Meanwhile, the noise effect is lightened.(6) A theoretical analysis is accomplished to prove that the polyvinylidene fluoride (PVDF) can be used to measure the structural dynamic strain. And the expressions about the structural strain and the tested PVDF electric charges for both 1D beam structures and 2D plate structures are deduced theoretically. A PVDF-based dynamic strain measurement experiment is designed and accomplished. The experimental results confirm the strain measuring ability of PVDF.(7) Decades of PVDF patches are used to measure dynamic strains of both an aluminum beam and plate in different damaged cases. The experimental data are obtained for damage identification using the PSED-based method. The results are acceptable although there are some defects in the experiment, which also confirms the practical effectiveness of the PSED-based method.