The integrity of the data collected by structural health monitoring systems has a significant impact on structural damage detection and state assessment. The missing or abnormal segments and unacquired future segments can be supplemented through signal reconstruction and prediction models. This paper proposes two novel models toward these two tasks based on bidirectional echo state network, which can exploit both historical and future signal segments to improve accuracies. Adaptive combination coefficient is introduced to control the rate of error accumulation. The effectiveness and robustness of the proposed models are verified by cases of synchronized missing, long-term missing, and boundary effect. A hyperparameter study related to both reservoir and memory is conducted to generate optimal models with maximum processing abilities. An ARIMAX and improved Kalman filter-based preprocessing method is adopted to keep all useful information and provide optimal estimation of the true signal values. The proposed models also show high performance in generating the high-frequency components. The superiority of the proposed models is validated through the datasets measured from Canton Tower, both stationary signals under free vibration and non-stationary signals under earthquake being considered.
FarrarCRWordenK. An introduction to structural health monitoring. Phil Trans R Soc A2007; 365: 303–315.
2.
KoJMNiYQ. Technology developments in structural health monitoring of large-scale bridges. Eng Struct2005; 27: 1715–1725.
3.
NiYQXiaYLiaoWY, et al. Technology innovation in developing the structural health monitoring system for Guangzhou New TV Tower. Struct Control Health Monit2009; 16: 73–98.
4.
ZhangQLYangBLiuT, et al. Structural health monitoring of Shanghai Tower considering time-dependent effects. Int J High-Rise Build2015; 4(1): 39–44.
5.
LiuTYangBZhangQL. Health monitoring system developed for Tianjin 117 high-rise building. J Aerospace Eng2017; 30(2): B4016004.
6.
OuJLiH. Structural health monitoring in mainland China: review and future trends. Struct Health Monit2010; 9(3): 219–231.
7.
OstachowiczWSomanRMalinowskiP. Optimization of sensor placement for structural health monitoring: a review. Struct Health Monitor2019; 18(3): 963–988.
8.
ZhangBYNiYQ. An inversed greedy method for information-based optimal sensor placement on bridges. In: Proceedings of the Structural Health Monitoring, Stanford, 2019, pp. 2825–2832.
9.
ZhangQLSunSYYangB, et al. Real-time structural health monitoring system based on streaming data. Smart Struct Syst2021; 28(2): 275–287.
10.
YangBPanLCZhuHT, et al. Spatiotemporal correlation analysis of the dynamic response of supertall buildings under ambient wind conditions. Struct. Design Tall Spec Build2022; 31(4): e1914.
11.
MagalhaesFCunhaACaetanoE. Vibration based structural health monitoring of an arch bridge: from automated OMA to damage detection. Mech Syst Signal Process2012; 28: 212–228.
12.
NiYQWangYWZhangC. A Bayesian approach for condition assessment and damage alarm of bridge expansion joints using long-term structural health monitoring data. Eng Struct2020; 212: 110520.
13.
MaoJWangHSpencerBFJr.Toward data anomaly detection for automated structural health monitoring: exploiting generative adversarial nets and autoencoders. Struct Health Monit2021; 20(4): 1609–1626.
14.
ZhangZSunC. Structural damage identification via physics-guided machine learning: a methodology integrating pattern recognition with finite element model updating. Struct Health Monit2021; 20(4): 1675–1688.
15.
MakridakisS. Time series prediction: forecasting the future and understanding the past. Int J Forecasting1994; 10(3): 463–466.
16.
BarretoGA. Time series prediction with the self-organizing map: a review. In: KacprzykJ (ed.) 2001Perspectives of neural-symbolic integration. 2007, pp. 135–158. Poland: Springer.
17.
HanZYZhaoJLeungH, et al. A review of deep learning models for time series prediction. IEEE Sens J2021; 21(6): 7833–7848.
18.
BoxGEJenkinsGMReinselGC, et al. Time series analysis: forecasting and control. Hoboken: John Wiley & Sons, 2015.
19.
LiuCHHoiSZhaoPL, et al. Online ARIMA algorithms for time series prediction. In: Proc. AAAI Conf. on artificial intelligence2016, Phoenix, Arizona, Karviná, Czech Republic, pp. 1867–1873.
20.
PeterDSilviaP. ARIMA vs. ARIMAX–which approach is better to analyze and forecast macroeconomic time series. In: Proc. Int. Conf. Mathematical methods in Economics2012, pp. 136–140.
21.
KalmanRE. A new approach to linear filtering and prediction problems. J Fluids Eng1960; 82(1): 35–45.
22.
JazwinskiAH. Stochastic processes and filtering theory. New York: Dover Publications, 2007.
23.
HoshiyaMSaitoE. Structural identification by extended Kalman filter. J Eng Mech1984; 110(12): 1757–1770.
24.
JooTWKimSB. Time series forecasting based on wavelet filtering. Expert Syst Appl2015; 42: 3868–3874.
25.
HearstMADumaisSTOsunaE, et al. Support vector machines. Intell Syst Appl1998; 13(4): 18–28.
26.
KimKJ. Financial time series forecasting using support vector machines. Neurocomputing2003; 55(1–2): 307–319.
27.
MaMCZioELinsID, et al. Failure and reliability prediction by support vector machines regression of time series data. Reliab Eng Syst Saf2011; 96: 1527–1534.
28.
ElmanJL. Finding structure in time. Cognit Sci1990; 14: 179–211.
29.
BengioYSimardPFrasconiP. Learning long-term dependencies with gradient descent is difficult. IEEE Trans Neural Netw1994; 5(2): 157–166.
30.
HochreiterSBengioYFrasconiP, et al. Gradient flow in recurrent nets: the difficulty of learning long-term dependencies. IEEE Press.
31.
SherstinskyA. Fundamentals of recurrent neural network (RNN) and long short-term memory (LSTM) network. Phys D: Nonlinear Phenomena2020; 404: 132306.
32.
HochreiterSSchmidhuberJ. Long short-term memory. Neural Comput1997; 9(8): 1735–1780.
33.
LiYTBaoTFGongJ, et al. The prediction of dam displacement time series using STL, extra-trees, and stacked LSTM neural network. IEEE Access2020; 8: 94440–94452.
34.
ChoKMerrienboerBGulcehreC, et al. Learning phrase representations using RNN encoder-decoder for statistical machine translation. In: Proceedings of the Empirical Methods in Natural Language Process2014, 1406.1078.
35.
JagerHHaasH. Harnessing nonlinearity: predicting chaotic systems and saving energy in wireless communication. Science2004; 304(5667): 78–80.
36.
JagerH. The “echo state” approach to analyzing and training recurrent neural networks-with an erratum note. Technical Report GMD Report 148, 2001.
37.
JiangJLaiYC. Model-free prediction of spatiotemporal dynamical systems with recurrent neural networks: role of network spectral radius. Phys Rev Res2019; 1(3): 033056.
38.
HajnalMALorinczA. Critical echo state networks. In: Proc. Int. Conf. Artificial Neural Networks, Athens, Greece, 10–14 September 2006, pp.658–667. Berlin, Heidelberg: Springer.
39.
RodanATinoP. Minimum complexity echo state network. IEEE Trans Neural Networks2010; 22(1): 131–144.
40.
LargerLFuentesABMartineghiR. High-speed photonic reservoir computing using a time-delay-based architecture Million words per second classification. Phys Rev X2017; 7(1): 011015.
41.
HeYYZhangLKChenZS, et al. A framework of structural damage detection for civil structures using a combined multi-scale convolutional neural network and echo state network. Eng Comput2022: 39: 1–19.
42.
ObstO. Distributed fault detection in sensor networks using a recurrent neural network. Neural Processing Lett2014; 40(3): 261–273.
43.
AntoneloEACamponogaraEFossB. Echo state networks for data-driven downhole pressure estimation in gas-lift oil wells. Neuron Networks2017; 85: 106–117.
44.
ZhaoRYanRQWangJJ, et al. Learning to monitor machine health with convolutional bi-directional LSTM networks. Sensors2017; 17(2): 273.
45.
HuangFWZhengWNGuoWZ, et al. Estimating missing data for sparsely sensed time series with exogenous variables using bidirectional-feedback echo state networks. CCF Trans Pervasive Comput Interact2022; 5: 1–19.
46.
WangQWangLQLiuY, et al. Time series prediction with incomplete dataset based on deep bidirectional echo state network. IEEE Access2019; 7: 152533-052544.
47.
PathakJHuntBGirvanM, et al. Model-free prediction of large spatiotemporally chaotic systems from data: a reservoir computing approach. Phys Rev Lett2018; 120(2): 024102.
48.
GallicchioCMicheliAPedrelliL. Deep reservoir computing: a critical experimental analysis. Neurocomputing2017; 268: 87–99.
49.
XueYBYangLHaykinS. Decoupled echo state networks with lateral inhibition. Neural Networks2007; 20(3): 365–376.
50.
XuDWWangYDJiaLM, et al. Real-time road traffic state prediction based on ARIMA and Kalman filter. Front Inform Technol Electronic Eng2017; 18: 287–302.
51.
ErazoKSenDNagarajaiahS, et al. Vibration-based structural health monitoring under changing environmental conditions using Kalman filtering. Mech Syst Signal Process2019; 117(15): 1–15.
52.
XinJZZhouJTYangS, et al. Bridge structure deformation prediction based on GNSS data using Kalman-ARIMA-GARCH model. Sensors2018; 18(1): 298.
53.
LukoseviciusMPopoviciDJagerH, et al. Time warping invariant echo state networks, International University Bremen, Technical Report, 2006. No. 2.
54.
JagerHLukoseviciusMPopoviciD, et al. Optimization and applications of echo state networks with leaky-integrator neurons. Neural Netw2007; 20(3): 335–352.
55.
LukoseviciusM. A practical guide to applying echo state networks. Neural Netw Tricks Trade2012; 27: 659–686.
ProkhorovDVFeldkarnpLATyukinIY. Adaptive behavior with fixed weights in RNN: an overview. In: Proceedings of the 2002 International Joint Conference on Neural Networks2002, Honolulu, HI, pp. 2018–2022.
58.
BarabasiALAlbertR. Emergence of scaling in random networks. Science1999; 286(5439): 509–512.
59.
KwiatkowskiDPhillipsPCBSchmidtP, et al. Testing the null hypothesis of stationarity against the alternative of a unit root: how sure are we that economic time series have a unit root. J Econom1992; 54(1–3): 159–178.
60.
FloresJHFEngelPMPintoRC. Autocorrelation and partial autocorrelation functions to improve neural networks models on univariate time series forecasting. In: The 2012 International Joint Conference on Neural Networks (IJCNN), Brisbane, QLD, Australia, 2012, pp. 1–8.
TanYKWangYLNiYQ, et al. Large-span bridge strain reconstruction based on bidirectional LSTM and ES. In: Proc. Int. Workshop of Structural Health Monitoring 2023, Stanford, p. 35637.
64.
ZhangFLNiYQNiYC, et al. Operational modal analysis of Canton Tower by a fast frequency domain Bayesian method. Smart Struct Syst2016; 17(2): 209–230.
