Damage Identification Study On Dynamic Response For Beam Members Based On Genetic Algorithm

During the entire service time of structures damage is bound to be accumulated due to such factors as fatigue, environmental erosion, and degradation of materials. The present damage identification technique relies on the acquisition of structure's global dynamic response signals, through which modal parameters are derived and then used to locate the damage position and determine the damage level. Despite the simplicity of this technique, difficulties are often encountered in its practical applications; problems arise when it comes to the noise measurement, determination of the number and locations of the sensors required, or the arrangement of the excitation positions. Therefore, combining this global response measurement, which can be employed to identify possible damage positions, with further local inspections for prospective structural component should be a viable and an efficient method. This thesis seeks to study beam member's damage identification by using local vibration technique, and the main work comes as below:First, a theoretical damage identification model is established by deriving the free vibration equation of a Timoshenko beam which is elastically constrained at both of its ends. When the mid element of the beam is damaged to some extent, the change of strain mode for the various nodes of the beam can be formulated by solving the standard matrix eigenvalue problem of the structure. The change of the element's average strain mode can then be interpolated from its nodal strain mode's changes. These formulations derived can easily be applied to calculate changes that occur to the non-damaged part of the beam due to the damage at the other part.Second, an ANSYS model is created to substitute the physical model, with both ends just as elastically constrained as the theoretical model above. The spring constant is used to simulate the constraints at the two ends while damage of the mid element is implemented by changing its Young's Modulus. The natural frequency, mode shapes and the element's average strain mode can be extracted from this ANSYS model.Finally, an objective function is formed by comparing the element's average strain mode calculated from the two models above and then minimizing the sum of the absolute values of the difference between these two groups of data. By further minimizing the absolute changes for the various elements'strain, a multi-objective optimization problem is presented to conduct damage identification for beam members. With the aid of genetic algorithm (GA), the thesis performs optimization identification for the damage coefficients when the beam member suffers from different level's damage and when the damage is located at its elastic constraints and the middle. And the results from different fitness functions are compared with each other. The damage identification technique formulated in this thesis can predict accurately, to some extent, the beam member's damage location and damage level, especially for the two cases of single-point damage and two-point damage.For example: the elastic constraint damage identification theory can be available on both ends of the seam of component; The damage identification theory of intermediate unit can be used for dangerous section of crane beams etc.