The problem of system identification and damage detection, an inverse problem of difficult treatment, lies currently on the algorithms to interpret the measured data without significant knowledge of the system a priori. Typical approaches require strong conditions on the number of sensors and actuators in the system in order to find a full order physical model of the structure. Using an evolutionary strategy, an optimization algorithm based on mechanisms inspired on natural evolution, this obstacle is overcome. In this thesis, this algorithm is presented for identification and damage detection of structures. An in-depth analysis of uniqueness of solutions for the identification problem of building-type structures is carried out, leading to novel conclusions regarding the minimum number of measurements to guarantee the uniqueness of solution. The identification algorithm is tested in conditions including limited data, output only data, and noise polluted signals without knowledge of mass and stiffness. Results are presented for the ASCE Benchmark problem.
展开▼
