Detection filter-based method with LMI technique for robust structural damage detection
Doctor of Philosophy thesis
Existence of structural damage in civil structures may greatly deteriorate the overall performance of the system or even lead to disastrous consequences. Therefore, structural damage detection and health monitoring are very important from life-safety and economic viewpoints. In this research, three new fault detection filter-based damage detection methods, including the equivalent static output feedback controller design method, eigenstructure assignment method and Hinfinity/H_ robust detection filter design method, are developed. Meanwhile, a new flexibility-based algorithm is proposed to estimate the extent of structural damage. Firstly, the problem of finding a stable detection filter is converted to the equivalent problem of finding a decentralized static output feedback controller. Then, the iterative Linear Matrix Inequalities (LMI) technique and Genetic Algorithm (GA) are, respectively, applied to find the stable detection filter gain. Although the above equivalent problem, iterative LMI technique and GA algorithm are not new, the idea of combing them together for structural damage detection is novel and has great potentials in solving robust damage detection filter gains. Secondly, a novel eigenstructure assignment method is developed and combined with LMI technique for robust structural damage detection. The obtained robust detection filter can reduce the noise effect on the isolated output residuals, which improves structural damage detection results. Thirdly, a novel damage detection method using H infinity/H_ concept and iterative LMI technique is proposed to find a stabilizing fault detection and isolation filter which not only bounds the Hinfinity norm of the transfer function from disturbances to the output residual, but also does not degrade the component of the output residual due to the fault. Fourthly, a two degree-of-freedom (DOF) system with semi-active independently variable stiffness (SAIVS) device is tested to verify the above proposed structural damage detection methods. The experiment results show that the proposed methods are effective and efficient. Finally, this thesis also develops a new flexibility-based numerical algorithm to detect structural damage. Based on the characteristics of flexibility matrix which is easy to obtain with enough accuracy experimentally, Gauss-Newton optimization method is used to find the optimal structural parameters. Comparing them with the healthy parameters, structural elements with and extent of damage can be determined.