Conference Paper

Frequency response based damage detection using principal component analysis

Dept. of Mech. Eng., Connecticut Univ., Storrs, CT, USA
DOI: 10.1109/ICIA.2005.1635122 Conference: Information Acquisition, 2005 IEEE International Conference on
Source: IEEE Xplore


In this paper we explore structural damage detection using frequency response signals and principal component analysis. While frequency responses are easy to measure especially in online damage detection applications, most of the associated detection methods are deterministic in nature and cannot deal with uncertainties and noise which are inevitable under practical situations. To tackle this issue and to develop a robust damage detection protocol, here we develop a feature extraction/de-noising methodology based on principal component analysis (PCA). The basic idea is to first establish a feature space of the intact structure response by using multiple measurements. Abnormal signature that is different from the baseline signature can then be identified and magnified after signal reconstruction using the intact structure features. Essentially, the directionality between an inspected signal and the baseline signal in the feature space is used as index of damage occurrence. A series of numerical analyses are performed to characterize the detection system sensitivity and robustness.

23 Reads
  • Source
    • "Available FRF data of the undamaged structure are also arranged in a separate matrix and mean values of each column are obtained. Even though PCA can reduce the dimensionality of the data set, the high dimensionality of the FRF dataset obtained from a complete observation can diminish the effectiveness of PCA [40]. Therefore, improvements for increasing the effectiveness of PCA are needed. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Pattern recognition is a promising approach for the identification of structural damage using measured dynamic data. Much of the research on pattern recognition has employed artificial neural networks (ANNs) and genetic algorithms as systematic ways of matching pattern features. The selection of a damage-sensitive and noise-insensitive pattern feature is important for all structural damage identification methods. Accordingly, a neural networks-based damage detection method using frequency response function (FRF) data is presented in this paper. This method can effectively consider uncertainties of measured data from which training patterns are generated. The proposed method reduces the dimension of the initial FRF data and transforms it into new damage indices and employs an ANN method for the actual damage localization and quantification using recognized damage patterns from the algorithm. In civil engineering applications, the measurement of dynamic response under field conditions always contains noise components from environmental factors. In order to evaluate the performance of the proposed strategy with noise polluted data, noise contaminated measurements are also introduced to the proposed algorithm. ANNs with optimal architecture give minimum training and testing errors and provide precise damage detection results. In order to maximize damage detection results, the optimal architecture of ANN is identified by defining the number of hidden layers and the number of neurons per hidden layer by a trial and error method. In real testing, the number of measurement points and the measurement locations to obtain the structure response are critical for damage detection. Therefore, optimal sensor placement to improve damage identification is also investigated herein. A finite element model of a two storey framed structure is used to train the neural network. It shows accurate performance and gives low error with simulated and noise-contaminated data for single and multiple damage cases. As a result, the proposed method can be used for structural health monitoring and damage detection, particularly for cases where the measurement data is very large. Furthermore, it is suggested that an optimal ANN architecture can detect damage occurrence with good accuracy and can provide damage quantification with reasonable accuracy under varying levels of damage.
    Engineering Structures 05/2014; 66:116–128. DOI:10.1016/j.engstruct.2014.01.044 · 1.84 Impact Factor
  • Source
    • "The primary sources of difficulties include measurement noise, modelling error, uncertainty of ambient conditions and incompleteness of measured data. An important aspect of damage identification research is therefore the discrimination of abnormal response variation due to damage, from normal response variation due to measurement noise, fluctuations of ambient conditions or operating uncertainty [12]. Artificial neural networks (ANNs), a form of artificial intelligence, have strong abilities to learn from experience, generalise from examples, and identify underlying information from noisy data. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paperpresentsastructuralhealthmonitoring(SHM)techniquethatutilisespattern changesinfrequencyresponsefunctions(FRFs)asinputparametersforasystemof artificialneuralnetworks(ANNs)toassessthestructuralconditionofastructure.To verifytheproposedmethod,itisappliedtonumericalandexperimentalmodelsofatwo- storey framedstructure,onwhichstructuraldamageisinducedbymemberconnectivity and masschanges,respectively.Forthenumericalstructure,simulatedtime-historydata are pollutedwithvariouslevelsofwhiteGaussiannoiseinordertorealisticallyrepresent field-testingconditions.Asadamageindicator,residualFRFsareused,whicharederived by calculatingthedifferencesinFRFdatabetweentheundamaged/baselinestructureand the structurewithchangedjointconditionsoraddedmass.Toobtainsuitablepatterns for neuralnetworktraining,principalcomponentanalysis(PCA)techniquesareadopted to reducethesizeoftheresidualFRFdataandtofilternoise.Ahierarchicalsystemof individualANNs,termednetworkensemble,isthentrainedtomapchangesinPCA- reducedresidualFRFstodamageconditions.Theresultsobtainedforbothdamage investigations,namelyjointdamageandmasschanges,demonstratethattheproposed SHM techniqueisaccurateandreliableinassessingtheconditionoftheteststructure numericallyandexperimentallybasedondirectFRFmeasurementsandnetwork ensembleanalysis.Fromtheoutcomesoftheindividualnetworks,itisfoundthatthe proposedhierarchicalnetworkensembleapproachishighlyefficientinfilteringpoor results ofunderperformingnetworksobtainedfrommeasurementlocationswithlow damagesensitivity.
    Journal of Sound and Vibration 08/2013; 336(16):3636. DOI:10.1016/j.jsv.2013.02.018 · 1.81 Impact Factor
  • Source
    • "In recent years, however, there is an emerging trend of using directly measured FRF data to detect structural damage (Huynh et al., 2005, Fang and Tang, 2005, Zang and Imregun, 2001, Ni et al., 2006, Li et al., 2009). Utilising FRFs to form a damage indicator may have several advantages: FRFs are closer to directly measured data and easy to obtain in real-time as they require only a small number of sensors and very little human involved processing (Fang and Tang, 2005). Measured FRF data are usually the most compact form of data obtained from vibrational testing, and hence, they provide an abundance of information on a structure's dynamic behaviour. "
    [Show abstract] [Hide abstract]
    ABSTRACT: This paper presents a vibration-based damage identification method that utilises damage fingerprints embedded in frequency response functions (FRFs) to identify location and severity of notch-type damage in a two-storey framed structure. The proposed method utilises artificial neural networks (ANNs) to map changes in FRFs to damage characteristics. To enhance damage fingerprints in FRF data, residual FRFs, which are differences in FRF data between the undamaged and the damaged structures, are used for ANN inputs. By adopting principal component analysis (PCA) techniques, the size of the residual FRF data is reduced in order to obtain suitable patterns for ANN inputs. A hierarchy of neural network ensembles is created to take advantage of individual characteristics of measurements from different locations. The method is applied to laboratory and numerical two-storey framed structures. A number of single notch-type damage scenarios of different locations and severities are investigated. To simulate field-testing conditions, numerically simulated data is polluted with white Gaussian noise of up to 10% noise-to-signal-ratio. The results from both numerical and experimental investigations show the proposed method is effective and robust for detecting notch-type damage in structures.
    Advances in Structural Engineering 05/2012; 15(5):743-757. DOI:10.1260/1369-4332.15.5.743 · 0.58 Impact Factor
Show more