Output-Only Damage Detection of Shear Building Structures Using an Autoregressive Model-Enhanced Optimal Subpattern Assignment Metric

Abstract

This paper proposes a novel output-only structural damage indicator by incorporating the pole-based optimal subpattern assignment distance with autoregressive models to localize and relatively assess the severity of damages for sheared structures. Autoregressive models can model dynamic systems well, while their model poles can represent the state of the dynamic systems. Structural damage generally causes changes in the dynamic characteristics (especially the natural frequency, mode shapes and damping ratio) of structures. Since the poles of the autoregressive models can solve the modal parameters of the structure, the poles have a close relationship with the modal parameters so that the changes in the poles of its autoregressive model reflect structural damages. Therefore, we can identify the damage by tracking the shifts in the dynamic system poles. The optimal subpattern assignment distance, which is the performance evaluator in multi-target tracking algorithms to measure the metric between true and estimated tracks, enables the construction of damage sensitive indicator from system poles using the Hungarian algorithm. The proposed approach has been validated with a five-story shear-building using numerical simulations and experimental verifications, which are subjected to excitations of white noise, El Centro earthquake and sinusoidal wave with frequencies sweeping, respectively; the results indicate that this approach can localize and quantify structural damages effectively in an output-only and data-driven way.

Performance Evaluation Metrics and Approaches for Target Tracking: A Survey

“…The choice of parameters was given in [ 53 ]. The OSPA distance has been used widely in the literature [ 44 , 54 , 55 , 56 ] and has better properties for the multitarget error evaluation than the Hausdorff metric. In addition, there are various improved methods based on the OSPA [ 57 ], which are be enumerated as follows: Generalized OSPA (GOSPA): In the GOSPA metric, we look for an optimal assignment between the truth targets and the estimated tracks, leaving missed and false targets unsigned [ 58 ].…”

“…The choice of parameters was given in [ 53 ]. The OSPA distance has been used widely in the literature [ 44 , 54 , 55 , 56 ] and has better properties for the multitarget error evaluation than the Hausdorff metric.…”

Prognostics and Health Management in Nuclear Power Plants: An Updated Method-Centric Review With Special Focus on Data-Driven Methods

“…These models are easy to implement with low computational cost, but their performance is heavily affected by the trend information of historical observations, which may be unreliable during incipient failure stage and for long-term forecasts (Baur et al, 2020). Recent examples of using regression-based prognostic models include Qian et al (2014) for bearing wear-out, Barraza-Barraza et al (2017) for crack growth in aluminum plates, Nguyen et al (2018) for NPP steam generator degradation, and Mei et al (2020) for shear building structural damage. In Markovian-based models, the degradation process is assumed to transform within a finite state space that satisfies the Markov (or memoryless) property.…”

An Unsupervised Learning Approach for Early Damage Detection by Time Series Analysis and Deep Neural Network to Deal with Output-Only (Big) Data

“…The former step is a signal processing strategy, which aims at extracting meaningful information (here called damage-sensitive features) from raw measured data (e.g., acceleration time histories), while the latter is a machine learning algorithm for analyzing and classifying the extracted features for early damage detection, localization and quantification [4][5][6][7]. Time series modeling is one of the powerful feature extraction methods, which is intended to fit a parametric representation (model) to raw measured data [8,9]. Coefficient-based and residual-based algorithms are two different feature extraction methods via time series modeling.…”