The structural health monitoring relies on the continuous observation of a dynamic system over time to identify its actual condition, detect abnormal behaviors, and predict future states. The regular changes in environmental factors have been reported as one of the main challenges for the application of structural health monitoring systems. These influences in the structural responses are in general nonlinear, affecting the damage-sensitive features in the most varied forms. The usual process to remove these normal changes is referred to as data normalization. In that regard, principal component analysis is probably the most studied algorithm in structural health monitoring, having numerous versions to learn strong nonlinear normal changes. However, in most cases, not all variability is properly accounted for via the existing nonlinear principal component analysis approaches, resulting in poor damage detection and quantification performances. In this article, a new paradigm based on deep principal component analysis, rooted in the deep learning field, is presented to overcome these limitations. This approach extracts the most salient underlying feature distributions by stacking multiple feedforward neural networks trained to learn an identity mapping of the input variables, where the network inputs are reproduced into the outputs. Similar to the traditional nonlinear principal component analysis–based approach, our approach identifies a nonlinear output-only model of an undamaged structure by comprising modal features into an internal bottleneck layer, which implicitly represents the independent environmental factors. The proposed technique is validated through the application on a progressively damaged prestressed concrete bridge and a three-span suspension bridge. The experimental results demonstrate that capturing the most slight nonlinear variations in the data can lead to improved data normalization and, consequently, better damage detection and quantification performances.
ZhouYLMaiaNMWahabMA.Damage detection using transmissibility compressed by principal component analysis enhanced with distance measure. J Vib Control2016; 24(10): 2001–2019.
2.
RoyKBhattacharyaBRay-ChaudhuriS.Arx model-based damage sensitive features for structural damage localization using output-only measurements. J Sound Vib2015; 349(Suppl. C): 99–122.
3.
FarrarCRWordenK.Chapter 7: damage-sensitive features. In: FarrarCRWordenK (eds) Structural health monitoring: a machine learning perspective. Chichester: John Wiley & Sons, Ltd, 2012, pp. 161–243.
4.
SampaioRMaiaNRibeiroA, et al. Using the detection and relative damage quantification indicator (DRQ) with transmissibility. In: GaribaldiLSuraceCHolfordK, et al. (eds) Damage assessment of structures VII (Key Engineering Materials, Volume 347). Aedermannsdorf: Trans Tech Publications, 2007, pp. 455–460.
5.
SohnH.Effects of environmental and operational variability on structural health monitoring. Philos T R Soc A2007; 365(1851): 539–560.
6.
XiaYChenBWengS, et al. Temperature effect on vibration properties of civil structures: a literature review and case studies. J Civ Struct Health Monit2012; 2(1): 29–46.
7.
ReyndersEWurstenGDe RoeckG.Output-only structural health monitoring in changing environmental conditions by means of nonlinear system identification. Struct Health Monit2013; 13: 82–93.
8.
KullaaJ.Is temperature measurement essential in structural health monitoring? In: Proceedings of the 4th international workshop on structural health monitoring: from diagnostic & prognostics to structural health monitoring, Stanford, CA, 15–17 September 2003, pp. 717–724. DEStech Publications: Stanford, California, USA.
9.
MeruaneVHeylenW.Structural damage assessment under varying temperature conditions. Struct Health Monit2012; 11(3): 345–357.
10.
HosseinabadiHZNazariBAmirfattahiR, et al. Wavelet network approach for structural damage identification using guided ultrasonic waves. IEEE T Instrum Meas2014; 63(7): 1680–1692.
11.
VanMKangHJ.Wavelet kernel local fisher discriminant analysis with particle swarm optimization algorithm for bearing defect classification. IEEE T Instrum Meas2015; 64(12): 3588–3600.
12.
D’AngeloGRamponeS.Feature extraction and soft computing methods for aerospace structure defect classification. Measurement2016; 85: 192–209.
13.
SahooBMaityD.Damage assessment of structures using hybrid neuro-genetic algorithm. Appl Soft Comput2007; 7(1): 89–104.
14.
BeenaPGanguliR.Structural damage detection using fuzzy cognitive maps and Hebbian learning. Appl Soft Comput2011; 11(1): 1014–1020.
15.
KangFLiJJXuQ.Damage detection based on improved particle swarm optimization using vibration data. Appl Soft Comput2012; 12(8): 2329–2335.
16.
PeetersBMaeckJDe RoeckG.Vibration-based damage detection in civil engineering: excitation sources and temperature effects. Smart Mater Struct2001; 10: 518–527.
17.
MansonG. Identifying damage sensitive, environment insensitive features for damage detection. In: Proceedings of the 3rd international conference on identification in engineering systems, Swansea, 15–17 April 2002, pp. 187–197University of Wales Location: Swansea.
18.
CrossEJMansonGWordenK, et al. Features for damage detection with insensitivity to environmental and operational variations. P Roy Soc Lond A Mat2012; 468(2148): 4098–4122.
19.
CrossEKooKBrownjohnJ, et al. Long-term monitoring and data analysis of the Tamar bridge. Mech Syst Signal Pr2013; 35(1–2): 16–34.
20.
BrownjohnJ.Structural health monitoring of civil infrastructure. Philos T Roy Soc A2007; 365(1851): 589–622.
YanAMKerschenGBoePD, et al. Structural damage diagnosis under varying environmental conditions —part I: a linear analysis. Mech Syst Signal Pr2005; 19(4): 847–864.
23.
PurarjomandlangrudiAGhapanchiAHEsmalifalakM.A data mining approach for fault diagnosis: an application of anomaly detection algorithm. Measurement2014; 55: 343–352.
24.
XiangJZhongYGaoH.Rolling element bearing fault detection using PPCA and spectral kurtosis. Measurement2015; 75: 180–191.
25.
YanAMKerschenGBoePD, et al. Structural damage diagnosis under varying environmental conditions—part II: local PCA for non-linear cases. Mech Syst Signal Pr2005; 19(4): 865–880.
26.
BaldiPHornikK.Neural networks and principal component analysis: Learning from examples without local minima. Neural Networks1989; 2(1): 53–58.
27.
KarhunenJJoutsensaloJ.Generalizations of principal component analysis, optimization problems, and neural networks. Neural Networks1995; 8(4): 549–562.
28.
WordenK.Structural fault detection using a novelty measure. J Sound Vib1997; 201(1): 85–101.
29.
KramerMA.Nonlinear principal component analysis using auto-associative neural networks. Aiche J1991; 37(2): 233–243.
30.
DeraemaekerAWordenK.New trends in vibration based structural health monitoring. New York: Springer-Verlag Wien, 2010.
31.
FigueiredoEParkGFarrarCR, et al. Machine learning algorithms for damage detection under operational and environmental variability. Struct Health Monit2011; 10(6): 559–572.
32.
ZhouHNiYKoJ.Structural damage alarming using auto-associative neural network technique: exploration of environment-tolerant capacity and setup of alarming threshold. Mech Syst Signal Pr2011; 25(5): 1508–1526.
33.
OhCKSohnHBaeIH.Statistical novelty detection within the Yeongjong suspension bridge under environmental and operational variations. Smart Mater Struct2009; 147(18): 125022–125029.
34.
YuqingZBingtaoSFengpingL, et al. NC machine tools fault diagnosis based on kernel PCA and k-nearest neighbor using vibration signals. Shock Vib2015; 2015: 139217.
35.
GoodfellowIBengioYCourvilleA.Deep Learning. Cambridge, MA: MIT Press, 2016.
36.
SchmidhuberJ.Deep learning in neural networks: an overview. Neural Networks2015; 61: 85–117.
37.
BengioYLamblinPPopoviciD, et al. Greedy layer-wise training of deep networks. In SchölkopfPBPlattJCHoffmanT (eds) Advances in neural information processing systems 19. Cambridge, MA: MIT Press, 2007, pp. 153–160.
HintonGESalakhutdinovRR.Reducing the dimensionality of data with neural networks. Science2006; 313(5786): 504–507.
40.
LuttrellSP.A unified theory of density models and auto-encoders. DERA technical report, DERA, Cambridge, 31October1997.
41.
BaldiP. Autoencoders, unsupervised learning and deep architectures. In: Proceedings of the 2011 international conference on unsupervised and transfer learning workshop, Irvine, CA, 2 July 2011Vol. 27, pp. 37–50JMLR W&P: Bellevue, Washington.
42.
Van der MaatenLPostmaEvan den HerikJ. Dimensionality reduction: a comparative review. Technical report, Tilburg University, Tilburg, 26October2009.
43.
VincentPLarochelleHBengioY, et al. Extracting and composing robust features with denoising autoencoders. In: Proceedings of the 25th international conference on machine learning, Helsinki, 5–9 July 2008, pp. 1096–1103. New York: ACM.
44.
DeMersDCottrellGW.Non-linear dimensionality reduction. In: HansonSJCowanJDGilesCL (eds) Advances in neural information processing systems 5. San Francisco, CA: Morgan Kaufmann Publishers Inc., 1993, pp. 580–587.
45.
RanzatoMAPoultneyCChopraS, et al. Efficient learning of sparse representations with an energy-based model. In: SchölkopfPBPlattJCHoffmanT (eds) Advances in neural information processing systems 19. Cambridge, MA: MIT Press, 2007, pp. 1137–1144.
46.
AndersonTW (ed.). An introduction to multivariate statistical analysis. New York: Wiley, 1984.
47.
KungSDiamantarasKITaurJ.Adaptive principal component extraction (APEX) and applications. IEEE T Signal Proces1994; 42(5): 1202–1217.
48.
SantosADFFigueiredoESilvaMFM, et al. Applicability of linear and nonlinear principal component analysis for damage detection. In: IEEE international instrumentation and measurement technology conference (I2MTC) proceedings, Pisa, 11–14 May 2015, pp. 869–875. New York: IEEE.
49.
BengioY.Learning deep architectures for AI. Found Trends Mach Learn2009; 2(1): 1–127.
50.
PeetersBRoeckGD.Reference-based stochastic subspace identification for output-only modal analysis. Mech Syst Signal Pr1999; 13(6): 855–878.
51.
FigueiredoERaduLWordenK, et al. A Bayesian approach based on a Markov-chain Monte Carlo method for damage detection under unknown sources of variability. Eng Struct2014; 80: 1–10.
52.
PeetersBDe RoeckG.One-year monitoring of the Z24-bridge: environmental effects versus damage events. Earthq Eng Struct D2001; 30(2): 149–171.
53.
SantosASilvaMSantosR, et al. A global expectation-maximization based on memetic swarm optimization for structural damage detection. Struct Health Monit2016; 15(5): 610–625.
54.
FigueiredoECrossE.Linear approaches to modeling nonlinearities in long-term monitoring of bridges. J Civ Struct Health Monit2013; 3(3): 187–194.
55.
FigueiredoERaduLWestgateR, et al. Applicability of a Markov-chain Monte Carlo method for damage detection on data from the Z-24 and Tamar suspension bridges. In: Proceedings of the 6th European workshop on structural health monitoring, Dresden, 3–6 July 2012, pp. 747–754. Berlin: DGZfP.
56.
ShannonCE.A mathematical theory of communication. Bell Syst Tech J1948; 27: 623–656.
