Research On The Structural Damage Detection Methods Based On The Vibration Response

With the development of science and technology, the engineering structures aredeveloping toward larger scale and more complicate. Due to the long-term changingloading condition and the harsh working environment, the strength reduction and thefatigue effect tend to be detected in the structure, and the working performance of thestructure will be greatly affected. To ensure the stability and the safeness of thestructure, it’s important to perform the structure health monitoring and detect thestructure damage at an early stage. The damage detection method based on thevibration testing is one of the most effective damage detection methods, which iswidely studied and applied. When the damage occurs, the damage information will becontained in the vibration responses, and the key to the damage detection method is toextract the effective damage indicator from the vibration responses.The modal parameters of the structure can be estimated using the vibrationresponse at different sensor, and the structural damage can be detected by comparingthe modal parameters between the normal structure and the damaged structure.However, the estimation of the modal parameters usually needs the excitation forceor presumably to be white noise. So it’s not fit for the structure damage detectionunder unknown excitation. On the other hand, the damage indicator can be extractedby studying the dynamic model of the structure and figuring out the damage effectsto the structure parameters, which needs a precise numerical model, so it’s notapplicable to the complex structure. In conclusion, it’s necessary to study the way ofextracting damage indicator directly through the vibration response, so as to detectstructural damage under unknown excitation. Supported by National Natural ScienceFoundation (No.51375152), this dissertation concentrates on the research of thedamage indicator extraction method and the damage detection method based on thevibration response. The main research works of the dissertation are as follows:(1) Due to the problem of directly utilizing the change of the transmissibilityfunction as a damage indicator tends to be affected by measurement error and noises,a damage detection method based on the transmissibility function and the singularvalue entropy is proposed. Firstly, the phase space reconstruction method is used toanalysis the transmissibility function to extract further information. Then, the singularvalue entropy is calculated as the damage indicator based on the reconstructed transmissibility function matrix, which is used to detect the damage of the threestorey bookshelf structure. The experiment results show the effectiveness of theproposed method. On the other hand, a damage detection method based on themulti-scale transmissibility function and the relative grey moment entropy is proposed.The multi-scale transmissibility function is derived by combining the wavelettransform and the transmissibility function, which represents the time-frequencycharacteristic of the transmissibility function. Then the damage indicator is obtainedby calculating the grey moment of the multi-scale transmissibility function atdifferent scale ranges, which remains the same under different loading condition. Andthe damage pattern can be identified by calculating the relative entropy between thesample data and the experimente data. The experiment results show that the proposedmethod can identify the damage pattern of the structure under the unknown andvarying loading condition.(2) Aiming to select proper damage parameters, a damage detection methodbased on the local mean decomposition method and the principal component analysisis proposed. The local mean decomposition method is used to decompose thevibration response into a series of PF components, and the characteristic parametersare calculated based on the PF components to assemble the characteristic parametermatrix. Then, the principal component analysis is applied to extract the principalcomponent, and the Euclidean distance is used to distinguish different damage pattern.The experiment results show that the proposed method can distinguish differentdamage pattern of the gear experiment structure. On the other hand, a damagedetection method based on the WPD-LMD method and the relative permutationentropy is proposed to deal with the deficiency of the LMD method, such as the lowenergy component cannot be extracted. The wavelet packet decomposition method isfirstly used to decompose the vibration response into a series of narrow bandcomponent, and then the LMD method is applied to the narrow band component. Theeffectiveness of the WPD-LMD is proved on a simulation signal. The damageindicator is obtained by calculating the permutation entropy, and the damage patterncan be identified by the relative permutation entropy between the test data and thesample data. The experimental results show the effectiveness of the method.(3) Aiming at the deficiency of the adaptive time-frequency method(the EMDand LMD), such as the end effect problem and the mode mixing problem, a newadaptive time-frequency method—ASTFA method is studied and applied to thevibration signal processing.A comparison is made between the ASTFA method and the EMD method to show the superiority of the ASTFA. The modal parameters areestimated and a damage index is proposed based on the instantaneous frequency andthe instantaneous energy of the component signal. The analysis results of the ASCEstructure and the free beam experiment show that the ASTFA method can be appliedto the structural damage detection. On the other hand, a damage identification methodbased on the transmissibility function and the ASTFA method is proposed. TheASTFA method is applied to decompose the time domain signal of the transmissibilityfunction, obtaining a series of Intrinsic mode function, and the cosine similarity of theIMF functions for the normal condition and the damage situation is calculated toselect the proper IMF component. The damage index is obtained by calculating thesummation of the instantaneous frequency and the energy moment. The experimentalresults show the effectiveness of the proposed method in damage identification.(4) To extract the damage indicator representing full-scale property of thestructure, a damage detection method based on the singular value decomposition andthe proper orthogonal decomposition is proposed. The singular value decomposition isfirstly applied to the structural vibration power spectral matrix to identify the modalfrequencies, and then the correlation matrix can be calculated at each modal frequency.The proper orthogonal decomposition method is applied on the correlation matrix, sothat the proper orthogonal mode can be obtained, which converges to the normalstructural mode. The proper orthogonal modes can be used to construct a damagelocating vector, and the damage can be located through the stress distribution of eachelements. The experimental results show that the proposed method can detect andlocate the damage effectively.