Improved structural damage identification using random decrement signatures: Application to FEM data

Abstract

This article describes detection of structural damage using statistical properties of randomdec signatures. Proposed technique is model free and does not require input measurements. The technique is evaluated using acceleration data obtained from a finite element model of a frame-like structure. Damage in the model is represented by a structural member with nonlinear stiffness characteristic due to opening and closing cracks. The results suggest that reliability of damage detection would depend on the orientation of the crack and the richness of the response. described in previously cited publications [1][2][3][4] were developed with the assumption of linearity of the structures and the changes that occur in them due to degradation. However, this is not necessarily the case in reality since structural damage introduces nonlinearity [5]. For example, beams containing breathing fatigue cracks would have displacement-dependent stiffness and are modeled as structures with bilinear stiffness characteristics [6]. Therefore, introduction of nonlinearities and other nonlinear phenomena could be exploited for structural health monitoring (SHM) purposes.Recently, some researchers started developing SHM methods that utilize nonlinear phenomena. Hunter [7] used an ARMA-based time series approach to characterize a system with bilinear stiffness characteristics. Different response regions were described by separate linear models. The authors were able to detect changes in the identified natural frequencies and mode shapes based on the obtained models. Another technique developed by Epureanu and Yin [8] exploited changes in the probability density functions of sampled attractors of an aeroelastic system. Adams and Nataraju [9] and Brown and Adams [10] proposed the use of nonlinear loworder state space models for damage variables.Nichols et al. [11,12] investigated the use of attractor dimension as a feature for SHM and compared different metrics reflecting attractor dimension. Moniz et al. [13] and Todd et al. [14] monitored the changes in the attractor formed from exciting the system by the output of a Lorenz oscillator. The assumption is that the system in a healthy state is a linear filter, although changes in the system (being linear or not) would cause the changes in the attractor.Chen et al.[15] studied degradation of reinforced concrete beams and discovered that their dynamic behavior was close to that of Duffing oscillator with softening stiffness. The authors developed a damage indicator based on the difference between the resonant frequencies obtained from dynamic tests with increasing and decreasing excitation frequencies. Wang et al.[16] defined a damage indicator based on coherence function that is sensitive to the onset of nonlinearity. The authors successfully detected damage in a steel frame. Thus, nonlinear response features can be useful indicators of damage.The goal of this work is to demonstrate a new SHM method based on changes in statistical properties of randomdec signatures caused by the onset of nonlin...

Localization of nonlinear damage using state-space-based predictions under stochastic excitation

“…Theoretically, the ERD technique can only applied to deal with linear signals/damages, but it has been successfully extended for nonlinear systems by previous researches [17,18], and promising results have obtained from the numerical simulation early in this paper. Capabilities of identifying nonlinear damages of the proposed method are further investigated using a more complex and realistic structure.…”

“…The original random decrement (RD) technique was introduced by Henry Cole in the early 1970s for estimating damping in aerospace structures [16]. Generally, given initial displacement, velocity, and random excitation, the "randomdec" signatures of the system can be extracted from the measured response x(t) using RD function δ(p) estimated as [17]:…”

“…The input force f(t) to the system is applied on the right hand side of the structure. To consider both linear and nonlinear damages, the spring element k2 is tuned in two ways: linear stiffness degradation and nonlinear change to the stiffness governed by , (17) where superscript d represents damage, and are linear and nonlinear damage severity coefficient respectively, and are the displacement at mass 1 and 2. All damage scenarios are listed in Table 1.…”

Damage identification of beam-like structures with contiguous and distributed damage

“…. Shiryayev and Slater described detection of structural damage in the form of an opening up and closing cracks through random decrement signature technique. Another combined static–dynamic system identification study using measured damage signature and predicted damage signature is through the work of Wang et al .…”

Detection of Fatigue Cracks Using Random Decrement Signatures

“…In particular, the feature studied here represents the change in the mean values of the signatures obtained between neighboring degrees of freedom (DOFs). In previous work [18], this method has been applied to the data obtained from the finite-element model of a frame-like structure, where the damaged structural element was modeled as a beam with nonlinear stiffness characteristic. The study showed that the technique allowed detection of crack type damage even in the presence of significant additive noise in the measurement data.…”