Data Normalization and Anomaly Detection in a Steel Plate-Girder Bridge Using LSTM

Abstract

: Modal properties are recognized as indicators reflecting structural condition in structural health monitoring (SHM). However, changing environmental and operational variables (EOVs) cause variability in the identified modal parameters and subsequently obscure damage effects. To address the issue caused by the EOVs-related variability, this study investigated the variability of modal frequencies in long-term SHM of a steel plate-girder bridge. A Bayesian fast Fourier transform (FFT) method was used for operational modal analysis in a probabilistic viewpoint. Bayesian linear regression (BLR) and Gaussian process regression (GPR) models were utilized to capture the variability in the identified most probable values (MPVs) of modal frequencies as temperature-driven models, and the limitation of these models for data normalization with latent EOVs was discussed. To overcome the interference of latent EOVs indirectly, a long short-term memory (LSTM) network was established to trace the variability as an autocorrelated process, with a traditional seasonal autoregressive integrated moving average (SARIMA) model as a benchmark. Finally, an anomaly detection method based on residuals of one-step ahead predictions by LSTM was proposed associating with the Mann-Whitney U test.

Eliminating environmental and operational effects on structural modal frequency: A comprehensive review

“…32 Besides, Bayesian linear regression (BLR) and Bayesian dynamic linear model (BDLM) can be utilized as temperature-driven models to capture the modal variability. 33,34 In view of the nonuniform temperature field distribution for large structures, an MLR model with the input of temperature or temperature gradient data of multiple measuring points is utilized and yields satisfactory accuracy in frequency prediction. 35 Sohn et al 22 estimated MLR using the least mean squares error minimization to forecast the temperature gradient-induced frequency variation; Xu et al 36 combined finite element model and MLR to consider the effects of temperature differences between bridge members and temperature gradient of girder and tower.…”

“…Besides, other machine learning-based or deep learning-based method can be introduced into the correlation modeling, such as the quantile random forest (QRF), ensemble learning, the long short-term memory (LSTM) applied for temperature-cable tension relation modeling, and the bidirectional LSTM (BiLSTM) for vertical deflection-cable tension relation modeling. 33,53 In addition, association rule learning (ARL) combined with traffic light was utilized to predict future bridge frequencies according to temperature measurements and detect anomalies. 54 A genetic programming method was applied to construct a predictor and capture temperature changes using the modal frequency as an independent variable.…”

“…This temperature‐driven model assumes that modal frequency is strongly linearly correlated with temperature variation 32 . Besides, Bayesian linear regression (BLR) and Bayesian dynamic linear model (BDLM) can be utilized as temperature‐driven models to capture the modal variability 33,34 …”

“…To consider the non‐Gaussianity of data distribution and the nonlinearity between temperature and modal frequency, Bayesian mixture of experts (MoE) model was developed to automatically switch among different regimes represented by experts 51,52 ; Wang et al 30 developed local PLSR for frequency refinement modeling in each linear region. Besides, other machine learning‐based or deep learning‐based method can be introduced into the correlation modeling, such as the quantile random forest (QRF), ensemble learning, the long short‐term memory (LSTM) applied for temperature–cable tension relation modeling, and the bidirectional LSTM (BiLSTM) for vertical deflection–cable tension relation modeling 33,53 …”

Early Anomaly Warning of Environment-Induced Bridge Modal Variability through Localized Principal Component Differences

“…Bridge structures are easily subjected to performance degradation during service under the combined effect of environmental corrosion, traffic load and natural disasters. To avoid catastrophic accidents caused by structural failure and timely identification of the occurrence of damage, the status (normal or damaged) of a structure should be effectively monitored (Kim et al 2018;Figueiredo 2019;Jiang et al 2021). In recent decades, with the rapid increase in structural health monitoring (SHM) systems installed on bridges, vibration-based damage warning methods have attracted extensive attention.…”