A Study On Information Entropy Theory Applied In Vibration-based Damage Identification Of Concrete Structures

In this thesis, based on a comprehensive review of structural damage identification methods, information entropy theory is proposed to partly solve some key problems existed in the field of vibration-based damage identification of concrete structure, and provide supporting criterions for concrete structual health monitoring.Based on information entropy theory, take auto-power spectrum entropy as an example, the construction method of entropy index functions which are suitable for concrete structural damage identification is proposed. The validity of auto-power spectrum entropy is experimentally testified by two different simply supported reinforced concrete beams. When applied to damage identification, the similarities and differences between auto-power spectrum entropy and power spectrum entropy are comparatively studied; the advantage of proposed method is demonstrated by experimental research. Noise immunity of auto-power spectrum entropy is investigated by adding various intensity of noise into original vibration signals.The programming method of approximate entropy (ApEn) theory is improved to enhance its computational efficiency. Decay signal is divided into various categories via auto power spectral density function, different ApEn methods are proposed according to different signal patterns. Cross-correlation function is applied to transform irregular decay signal into regular cross-correlation time series, the feasibility of introducing ApEn into calculating decay signal is validated by numerical simulation and experimental method. The experimental research shows that, ApEn theory could effectively detect damage in concrete structure with a strong sensitivity and noise immunity.Both original and improved linearized transfer entropy theory are introduced to detect damage in beam structures by acceleration signals. Simply supported concrete beams are analyzed using Abaqus, and the bending vibration of the beams only exits in its principal plane is assumed during the analysis. Elastic damage in the concrete beams is simulated via reducing the elastic modulus in a specified element. The linearized transfer entropy method is validated by kernel density estimation method based on the acceleration signals of the beam. By defining an appropriate damage index, improved linearized transfer entropy is proved to be very sensitive and robust when applied to damage identification of concrete structures.A numerical study of Duffing soft spring oscillator subjected to Gaussian excitation is analyzed using Matlab, the reasonability of linearized transfer entropy and kernel density estimation transfer entropy is verified by analytical solution of transfer entropy. Various degrees of nonlinear damage are simulated by adjusting nonlinear factor of the system, transfer entropy combined with surrogate data method are used to detect nonlinear damage, the result shows that this method can effectively capture the nonlinear characteristics of the system. A simply supported concrete beam with crack is analyzed using Abaqus, the nonlinear damage degree is simulated by the length of crack. Transfer entropy combined with surrogate data method is used to identify the nonlinear damage of continuous system.The effects of using transfer entropy and improved surrogate data method to recognize nonlinear damage of concrete structures are experimentally verified by three different simply supported reinforced concrete beams. The damage is quantified by defining appropriate damage index; intact beam and defective beam are used to verify this method has the ability to identify different nonlinear damage patterns. The robustness of transfer entropy combined with surrogate data method is testified from the perspective of sampling frequency, noise immunity, and excitation method.On the basis of theoretical analysis, numerical simulation and experimental studies, this thesis verifies the reliabilities and advantages of several different information entropy indexes when applied to dynamic damage identification of concrete structures, and aims at providing a new and feasible research idea for concrete structural damage identification methods.