Damage identification on large-scale truss structure using sparse vector recovery
Conference Year
January 2019
Abstract
In recent years, finite element model updating methods which target sparse solutions have been suggested as a means to quantify and locate spatially local damage from highly incomplete modal information. Despite the growing number of sparsity approaches to damage identification, most proposed methods were tested on numerical simulations and experiments comprised of simple model structures. In this paper, the author investigates the application of sparse vector recovery methods on a large-scale and complex aluminum structure strictly using shifts in the identified natural frequencies. Three sparse vector recovery algorithms are considered: l1-norm optimization, non-negative constrained least squares, and a novel approach l0-norm optimization. The algorithms are tested on vibration data taken from a 17.4(m) long three dimensional aluminum highway sign support truss in various damage states.
Primary Faculty Mentor Name
Eric Hernandez
Status
Graduate
Student College
College of Engineering and Mathematical Sciences
Program/Major
Civil Engineering
Primary Research Category
Engineering & Physical Sciences
Damage identification on large-scale truss structure using sparse vector recovery
In recent years, finite element model updating methods which target sparse solutions have been suggested as a means to quantify and locate spatially local damage from highly incomplete modal information. Despite the growing number of sparsity approaches to damage identification, most proposed methods were tested on numerical simulations and experiments comprised of simple model structures. In this paper, the author investigates the application of sparse vector recovery methods on a large-scale and complex aluminum structure strictly using shifts in the identified natural frequencies. Three sparse vector recovery algorithms are considered: l1-norm optimization, non-negative constrained least squares, and a novel approach l0-norm optimization. The algorithms are tested on vibration data taken from a 17.4(m) long three dimensional aluminum highway sign support truss in various damage states.