Hybrid approach for damage detection in flexible structures

Abstract

Recently, various system identification approaches have been developed and applied for the detection of in flexible structures. In this paper, modal characteristics extracted from vibration tests are used with an original finite element model in an identification approach developed to combine the advantages of two classes of techniques: eigensensitivity and multiple-constraint matrix adjustment. Here, physical property parameters are adjusted, as with eigensensitivity techniques, but model matrix characteristics are employed, as with matrix adjustment methods. The performance of this hybrid technique is shown with its application to data from a flexible-truss laboratory experiment. HE ability to locate and assess in flexible structures is becoming increasingly important for improving the perfor- mance and life of these systems. Many focused (or local) approaches have been developed and evaluated for the purpose of lo- cation and assessment, including X-ray, optical, infrared, and ultra- sonic methods. Among the system (or global) methods currently in development, those that use vibration response and system identifi- cation have progressed considerably in the last five years.16 Here, system identification refers to the process of obtaining an updated model to match the measured response. The motivation behind a system identification approach for detection is to quantify the information contained in the response as much and as effectively as possible. In this work, to the structure refers to localized failure of a part of the structure. This failure can be a complete loss of capability in the part or a degradation of properties to an unaccept- able level. We assume this failure or degradation would primarily affect the stiffness properties and therefore the modal characteristics of the dynamic response of the structure. This work is focused on in truss structures, so the algorithms that follow are applied to locate and assess the total or partial loss of stiffness in single or multiple members of a truss. However, these algorithms are not restricted to truss applications by design and so are applicable to other structures as well. In general, a predamage finite element method (FEM) model is constructed and possibly adjusted to match the response of the un- damaged structure. The inertia properties of this model are assumed to be essentially consistent with those of the damaged structure de- fined above. Recent researchers have adopted a two-step detection process: first location, then assessment. For the first step, the use of damage vectors, as presented by Zimmerman and Kaouk,5 or residual force vectors, as presented by Ricles and Kosmatka,7 can locate damaged regions of the structure as represented in the FEM model. The second step is to assess the by algorithmi- cally comparing the modal characteristics of the predamage FEM model to the postdamage structural response, producing an adjusted FEM model with adjustments limited to the region or regions defined in the first step. In this paper, a method is presented for the assessment step that uses the modal characteristics extracted from a vibra- tion test in a strategy developed to combine the advantages of two widely used system identification approaches: eigensensitivity