This paper proposes a Deep Learning Enhanced Principal Component Analysis (PCA) approach for outlier detection to assess the structural condition of bridges. We employ partially explainable autoencoder architecture to replicate and enhance the data compression and reconstruction ability of PCA. The particularity of the method lies in the addition of residual connections to account for nonlinearities. We apply the proposed method to monitoring data obtained from two bridges under real operation conditions and compare the results before and after adding the residual connections. Results show that the addition of residual connections enhances the outlier detection ability of the network, allowing to detect lighter damages.
FarrarCWordenK. Structural health monitoring a machine learning perspective. ISBN 78-1-119-99433-6. Hoboken, NJ: Wiley, 2013. DOI: 10.1002/9781118443118.
2.
FriswellMI. Damage identification using inverse methods. Philos T R Soc A2007; 365(1851): 393–410. DOI: 10.1098/rsta.2006.1930.
3.
SunLShangZXiaY, et al.Review of bridge structural health monitoring aided by big data and artificial intelligence: from condition assessment to damage detection. J Struct Eng2020; 146(5): 04020073. DOI: 10.1061/(ASCE)ST.1943-541X.0002535.
4.
AzimiMEslamlouADPekcanG. Data-driven structural health monitoring and damage detection through deep learning: state-of-the-art review. Sensors2020; 20(10): 2778. DOI: 10.3390/s20102778.
5.
SalehiHBurgueñoR. Emerging artificial intelligence methods in structural engineering. Eng Struct2018; 171: 170–189. DOI: 10.1016/j.engstruct.2018.05.084.
6.
WordenKLaneA. Damage identification using support vector machines. Smart Mater Struct2001; 10: 540–547. DOI: 10.1088/0964-1726/10/3/317.
7.
DervilisNAntoniadouIBarthorpeRJ, et al.Robust methods for outlier detection and regression for SHM applications. Int J Sustain Mater Struct Syst2015; 2(1/2): 3. DOI: 10.1504/ijsmss.2015.078354.
8.
RytterA. Vibrational based inspection of civil engineering structures. PhD Thesis, University of Aalborg, Denmark, 1993. PhD.-Thesis defended publicly at the University of Aalborg, April 20, 1993 PDF for print: 206 pp.
9.
SmarslyKDragosKWiggenbrockJ. Machine learning techniques for structural health monitoring. 8th Eur Workshop Struct Health Monit2016; 2: 1522–1531.
10.
DackermannULiJSamaliB. Dynamic-based damage identification using neural network ensembles and damage index method. Adv Struct Eng2010; 13(6): 1001–1016. DOI: 10.1260/1369-4332.13.6.1001.
11.
GordanMRazakHAIsmailZ, et al.Recent developments in damage identification of structures using data mining. Latin Am J Sol Struct2017; 14(13): 2373–2401. DOI: 10.1590/1679-78254378.
12.
PozoFArrugaIMujicaLE, et al.Detection of structural changes through principal component analysis and multivariate statistical inference. Struct Health Monit2016; 15(2): 127–142. DOI: 10.1177/1475921715624504.
13.
JolliffeI. Principal component snalysis. ISBN 978-3-642-04898-2. Berlin, Heidelberg: Springer Berlin Heidelberg, 2011, p. 1094–1096. DOI: 10.1007/978-3-642-04898-2_455.
14.
ComanducciGMagalhãesFUbertiniF, et al.On vibration-based damage detection by multivariate statistical techniques: application to a long-span arch bridge. Struct Health Monit2016; 15(5): 505–524. DOI: 10.1177/1475921716650630.
15.
Soo Lon WahWChenYTElaminA, et al.Damage detection under temperature conditions using pca – an application to the z24 bridge. Proc Inst Civil Eng-Struct Build2020; 173: 1–13. DOI: 10.1680/jstbu.20.00024.
16.
Garcia-SanchezDFernandez-NavamuelASánchezDZ, et al.Bearing assessment tool for longitudinal bridge performance. J Civil Struct Health Monit2020; 10: 1–25. DOI: 10.1007/s13349-020-00432-1.
17.
García-MacíasEUbertiniF. Mova/moss: two integrated software solutions for comprehensive structural health monitoring of structures. Mech Syst Signal Process2020; 143: 106830, DOI: 10.1016/j.ymssp.2020.106830.
18.
MujicaLRodellarJGuemesA, et al.Pca based measures: Q-statistic and t2-statistic for assessing damages in structures. Proc 4th Eur Workshop Struct Health Monit2008; 1: 1088–1095.
19.
RainieriCMagalhaesF. Challenging aspects in removing the influence of environmental factors on modal parameter estimates. Proced Eng2017; 199: 2244–2249. DOI: 10.1016/j.proeng.2017.09.210. X International Conference on Structural Dynamics, EURODYN 2017.
20.
KramerMA. Nonlinear principal component analysis using autoassociative neural networks. AIChE J1991; 37(2): 233–243. DOI: 10.1002/aic.690370209.
HintonGESalakhutdinovRR. Reducing the dimensionality of data with neural networks. Science2006; 313: 504–507.
23.
BourlardHKampY. Auto-association by multilayer perceptrons and singular value decomposition. Biol Cybernetics1988; 59: 291–294. DOI: 10.1007/BF00332918.
24.
GoodfellowIBengioYCourvilleA. Deep learning. Cambridge: The MIT Press, 2016. ISBN 0262035618.
25.
SakuradaMYairiT. Anomaly detection using autoencoders with nonlinear dimensionality reduction. In: ACM International Conference Proceeding Series, Gold Coast, Australia, 4–11 December 2014. 2014; 02. DOI: 10.1145/2689746.2689747.
26.
NguyenVHMahowaldJGolinvalJ-C, et al.Damage detection in civil engineering structure considering temperature effect. Conference Proceedings of the Society for Experimental Mechanics Series, Orlando, FL, 2014; 4: 187–196. DOI: 10.1007/978-3-319-04546-7_22.
27.
GhoulemKKormiTAliNBH. Damage detection in nonlinear civil structures using kernel principal component analysis. Adv Struct Eng2020; 23(11): 2414–2430. DOI: 10.1177/1369433220913207.
28.
AngelovPSoaresE. Towards explainable deep neural networks (xDNN). Neural Networks2020; 130: 185–194. DOI: 10.1016/j.neunet.2020.07.010.
29.
MagalhãesFCunhaACaetanoE. Vibration based structural health monitoring of an arch bridge: from automated OMA to damage detection. Mech Syst Signal Process2012; 28: 212–228. DOI: 10.1016/j.ymssp.2011.06.011.
30.
RainieriCGargaroDFabbrocinoG, et al.Shaking table tests for the experimental verification of the effectiveness of an automated modal parameter monitoring system for existing bridges in seismic areas. Struct Control Health Monit2018; 25(7): e2165. DOI: 10.1002/stc.2165.
31.
CancelliALaflammeSAlipourA, et al.Vibration-based damage localization and quantification in a pretensioned concrete girder using stochastic subspace identification and particle swarm model updating. Struct Health Monit2019; 19: 147592171882001. DOI: 10.1177/1475921718820015.
32.
LimongelliMPGiordanoPF. Vibration-based damage indicators: a comparison based on information entropy. J Civil Struct Health Monit2020; 10: 251–266. DOI: 10.1007/s13349-020-00381-9.
33.
FathiALimongelliMP. Statistical vibration-based damage localization for the s101 bridge, flyover reibersdorf, Austria. Struct Infrastruct Eng2020; 0(0): 1–15. DOI: 10.1080/15732479.2020.1777570.
34.
FrankMDrikakisDCharissisV. Machine-learning methods for computational science and engineering. Computation2020; 8(1): 15–35. DOI: 10.3390/computation8010015.
35.
ChenJSatheSAggarwalC, et al.Outlier detection with autoencoder ensembles. In: Proceedings 17th SIAM Int Conference Data Mining, SDM, Houston, TX, 2017, pp. 90–98. DOI: 10.1137/1.9781611974973.11.
36.
OhDYYunID. Residual error based anomaly detection using auto-encoder in SMD machine sound. Sensors (Basel)2018; 18(5): 1–14. DOI: 10.3390/s18051308.
37.
LiLFangYWuJ, et al.Autoencoder based residual deep networks for robust regression prediction and spatiotemporal estimation. CoRR2018; (arXiv preprint arXiv:1812.11262).
38.
NiFZhangJNooriM. Deep learning for data anomaly detection and data compression of a long-span suspension bridge. Comp-Aided Civil Infrastruct Eng2019; 35: 685–700. DOI: 10.1111/mice.12528.
39.
MaXLinYNieZ, et al.Structural damage identification based on unsupervised feature-extraction via variational auto-encoder. Measurement2020; 160: 107811, DOI: 10.1016/j.measurement.2020.107811.
40.
XuBSongGMasriSF. Damage detection for a frame structure model using vibration displacement measurement. Struct Health Monit2012; 11(3): 281–292. DOI: 10.1177/1475921711430437.
da FonsecaAA. The infant henrique bridge over the river douro, in Porto, Portugal. Int Conf Arch Bridges2007; 5: 931–960.
43.
Adão da FonsecaABastosRO. Monitorização em fase de serviço do comportamento estrutural da Ponte Infante D. Henrique. Encontro Nacional de Betão Estrutural2004; Ø13 mm: 24–26.
