This paper presents an improved numerical tool for identification of damage in skeletal structures. The problem of identification has been formulated in the time domain within the framework of the Virtual Distortion Method (VDM). VDM generally belongs to fast structural reanalysis methods and can be applied to Structural Health Monitoring problems, among others. The major computational asset of VDM is the influence matrix, containing all the local-global inter-relations for a structure due to given perturbations e.g. initial strain or external force. A non-linear least squares problem with strains, entering the objective function, is the subject of consideration. Strains are used in order to have relatively smooth variations (compared to accelerations) of the analyzed signal in time. The change of stiffness is the design variable. Analytical gradients are implemented in the optimization code based on the Levenberg-Marquardt algorithm with some penalty function terms. The efficiency of the software tool is demonstrated for a numerical example of a 2D truss structure. A breakthrough in terms of computational time reduction has been observed compared to the previously used steepest-descent optimization. The presented software assumes the feasibility of reliable measurements of strains in time for real skeletal structures (e.g. truss bridges). Future research will include experimental verification of the idea with piezoelectric sensors acting as tensometers.
展开▼
