Structural Damage Identification Using Sensitivity-Enhancing Control And Statistical Analysis

With the continuous construction of large engineering structures in aerospace, civil, ocean, and mechanical engineering communities and the aging of existing structures in these communities, development of technology to monitor a structure and detect damage is becoming increasingly important. During the last 30 years, non-destructive examination (NDE)method has received considerable attention both in the academic and engineering communities. As the one kind of NDE vibration-based structural damage detection has become a hot research area because of its simplicity and minimum interaction with users and has promising applicability. However, with the development of research and acquirement of its practical application, many techniques of the vibration-based damage identification have shown some limitations.This thesis first presents a comprehensive summary and the state-of-the-art review on development of vibration-based structural damage detection. In the study, emphases are placed on how to enhance sensitivity of the damage indicator to damage and how to utilize time-domain response data of structure for detecting damage. The main works presented in the thesis are as follows:(1)The principle and feasibility of the sensitivity-enhancing feedback control for damage detection is further explored. Feedback control based on independent modal space control is first used to assign the pole of the system under detection intentionally. Then the prescribed characteristic frequencies of closed-loop system, which are more sensitive to damage, are obtained and further employed to constitute a sensitivity-enhanced damage indicator (SEDI). To overcome the effect of measurement noise on modal frequencies, a hypothesis test involving the t-test that utilizes the SEDI is employed to estimate the occurrence of damage. The combination of sensitivity-enhancing feedback control and statistical analysis is expected to improve the capability of the frequencies of the closed-loop system for identifying small damage and to lower the sensitivity to measurement noise.(2)A statistical pattern recognition technique is used for locating damage with the characteristic frequencies of the closed-loop system. Feedback control based on independent modal space control is used to assign the pole of the system under detection intentionally, and then the frequencies of the closed-loop system are used to construct feature vector. Based on perturbation theory, the feature vectors are normalized in order to eliminate the effect of damage extent on damage localization. Finally, mahalanobis distance of multivariate statistical analysis is used for locating damage.(3)A novel method using time-domain response data under random loading for detecting structural damage is proposed. A time series model with a fitting order is first constructed using the time domain response data with measurement noise. A sensitivity matrix consisting of the first differential of the autoregressive coefficients of the time series models with respect to the stiffness of the structural elements is then obtained. The locations and severities of the damage may be finally estimated by solving for the damage vector whose components are the damage degrees of the structural elements. A unique aspect of this method is that acceleration history data obtained from only one or a few sensors are required for detection and more feasibility for sensor arrangement is obtained accordingly. This advantage is helpful to reduce the difficulty and cost of testing of damage detection especially for large-scale complex engineering structures such as offshore platform structures.(4)The sensitivity of autoregressive coefficient to element stiffness is deduced, and it is concluded that assignment of pole of system under detection can be employed to enhance the sensitivity of autoregressive coefficient to element stiffness to improve the accuracy of damage detection. Principal component analysis is first carried out on all response time series of closed-loop system for data compression. A time series model with a fitting order is then constructed using the fist principal component. Finally, an X-bar control chart is constructed based on the mean value of autoregressive coefficient. The identification of damage occurrence is performed by monitoring the statically significant change of the control chart. Because only time-domain responses data are demanded and the sensitivity of autoregressive coefficient is enhanced by feedback control, the presented approach is efficient for the identification of early small damage and is very attractive for online structural monitoring system.(5)In order to demonstrate the effectiveness and feasibility of using control chart for damage detection, a beam structure with damage is tested in laboratory. Transfer function and modal frequencies of the damage and undamaged beams are measured. An analysis about the change of vibration characteristics of structures is carried out according to the measurement data. The process using control chart for the identification of damage occurrence is performed based on acceleration samples of the beam. High success rates are obtained.In this thesis, statistical approach combined with control theory are utilized for structural damage detection in an effort to enhance the sensitivity of damage feature indicator with measurement noise, and several key techniques and basic theories are studied. These research findings can be employed to supply novel ideas and approaches for improvement of robust ability of vibration-based damage identification, enhancement of the sensitivity of damage feature indicator to damage and the use of time-domain response data applied to damage detection.