A Wavelet based Damage Diagnosis Algorithm Using Principal Component Analysis

Size: px
Start display at page:

Download "A Wavelet based Damage Diagnosis Algorithm Using Principal Component Analysis"

Transcription

1 A Wavelet based Damage Diagnosis Algorithm Using Principal Component Analysis K. Krishnan Nair and Anne S. Kiremidian K. Krishnan Nair, Post Doctoral Student, Departments of Civil and Environmental Engineering & Economics Stanford University, Stanford, CA Tel.: (650) ; Fax: (650) ; Anne S. Kiremidian (CORRESPONDING AUTHOR) Professor, Department of Civil and Environmental Engineering, Stanford University, Stanford, CA Tel.: (650) ; Fax: (650) ; Figures: 9 Tables: 1 1

2 A Wavelet based Damage Diagnosis Algorithm Using Principal Component Analysis 1 K. Krishnan Nair and Anne S. Kiremidian Department of Civil and Environmental Engineering, Stanford University, CA {kknair,ask}@stanford.edu SUMMARY Earlier papers by the authors have shown the applicability of the damage sensitive feature based on the Haar and Morlet wavelet transform of the vibration signal. In the first part of the paper, a data preprocessing algorithm is developed to compare two vibration signals which enables optimal signal selection from a database of baseline signals. The second part of the paper describes the extraction of the damage sensitive feature vector as a function of the energies at the fifth, sixth and seventh dyadic scales of the vibration signal. Both data preprocessing and feature extraction steps involve the use of principal components analysis. The process of damage detection is automated using the k-means algorithm and the Gap statistic. A simple damage extent measure is also discussed. Finally, the migration of damage sensitive feature vectors with damage is illustrated for vibration signals obtained from the ASCE Benchmark Structure. The results indicate that the developed algorithm is able to consistently detect and quantify damage for the damage patterns specified by the ASCE Benchmark Experiment. Keywords: Structural safety; Damage detection; Statistical Analysis; Signal processing; Vibration analysis. 1 This research was partially supported by the John A. Blume Doctoral Fellowship and the NSF-CMMI Grant The authors would also like to acknowledge the ASCE Benchmark Committee for the data and MATLAB codes. The authors would also like to thank Haeyoung Noh for providing excellent research assistance. 2

3 1. INTRODUCTION For early detection and assessment of structural damage, it is critical to monitor the structural response data by installing sensors on the structure under question. Wireless sensor technology has been used for this purpose and has several advantages over their wired counterparts [1, 2]. Despite its advantages, wireless sensors require low power consumption which can be achieved by embedding the damage diagnosis algorithm at the sensing unit. Thus for embedding online damage detection algorithms at the sensor level, statistical signal processing techniques along with pattern classification schemes have been used [3, 4]. These algorithms depend on features or signatures obtained from the recorded structural response signals such as acceleration, strain or other data that change with the onset of damage. Subsequently, these features are classified using pattern classification schemes. Such algorithms involve the following steps: (i) the evaluation of a structure s operational environment which include loading conditions, temperature and humidity, (ii) the acquisition of structural response measurements and data preprocessing, (iii) the extraction of features that are sensitive to damage, and (iv) the development of statistical models for feature discrimination [3]. Previous papers by the authors have discussed the applicability of a damage sensitive feature based on the energies of the Haar and Morlet wavelet transforms of the vibration signal [5, 6]. In this paper, a damage detection algorithm based on this feature is presented. Unlike previous studies, the algorithm requires the creation of a database of normalized baseline signals and a data preprocessing algorithm is developed using principal components analysis of wavelet energies to obtain that baseline signal which is closest to the signal which is being analyzed. The damage sensitive feature is obtained by extracting the minimum variance obtained by the principal components analysis of the wavelet energies at the 5 th, 6 th and 7 th dyadic scales. The k-means algorithm estimates the mixture centers in a dataset and the gap statistic is used to determine the optimal number of clusters in the dataset. It is hypothesized that if more than one cluster is found while comparing the damage sensitive feature of the closest baseline signal and signal being analyzed, then a change has taken place and this change is most likely due to damage. 3

4 This paper is organized as follows: first, a methodology is developed for selecting the signal in the database that is closest to the new signal. This selection procedure is achieved using the lower singular values of the energies of the wavelet coefficients at the first dyadic scale. Next, the damage sensitive feature vector involving the principal components analyses of the energies of the wavelet coefficients at the fifth, sixth and seventh dyadic scales is presented. Damage detection using the k-means algorithm and the gap statistic is developed in the next section. Also, a simple Euclidean metric is developed as a damage extent measure. Finally, this algorithm is tested using datasets from the ASCE Benchmark Structure [7]. 2. OVERVIEW OF ALGORITHM This algorithm requires creating a database of baseline measurements (which can include accelerations and strains) and computing the coefficients of the continuous Daubechies wavelet of order four (DB4) at appropriate scales. Baseline measurements would include vibration and strain signals obtained from the undamaged structure (at various locations) under different operational conditions such as loading and environmental conditions. The DB4 wavelet is chosen because it is similar to the Morlet wavelet. Moreover it has a discrete wavelet counterpart which makes it particularly suitable for embedding the algorithm at the sensor level. Following this, principal components analysis is performed on the energies of the DB4 wavelet coefficients at the first dyadic scale. This step helps in obtaining the closest signal in the database, which describes the loading condition of the new signal. Once the closest baseline signal in the database is chosen, feature vectors are calculated using the DB4 wavelet coefficients at the fifth, sixth and seventh dyadic scales. Intuitively, the fifth, sixth and seventh dyadic scales correspond to lower frequency modes, where in general damage is observed. This is because the data used in this paper comes from the ambient vibrations; it is unlikely that the higher modes of vibration (corresponding to lower scales) will be excited. Finally the k-means algorithm and the gap statistic are used to discriminate between damaged and undamaged states in the structure. Damage is detected when the algorithm predicts more than one cluster in the feature vectors under question. The proposed detection algorithm is as follows: 4

5 i Obtain signals from an undamaged structure, from sensor i, denoted by x i (t) (i = 1,,P), where P is the number of sensors. Divide the signal x i (t) into segments / chunks of finite duration x i (t) ( = 1,,Q ), where Q is the number of segments. Standardize these signals to obtain a mean zero and standard deviation one signal. The database is populated with these standardized baseline signals. ii Compute the DB4 wavelet coefficients of the baseline signals at the first, fifth, sixth and seventh dyadic scales. Calculate the norm of the wavelet coefficients within a window size of d data points 2. These are represented by E [ E E E ] { 1 Q} baseline, = 1, baseline, 5, baseline, 6, baseline, E7, baseline,,..., (1) where E i,baseline, is the energy vector of the wavelet coefficients of the th baseline signal at the ith dyadic scale. It is noted that E i,baseline, (i = 1, 5, 6, 7; = 1,,Q) is a vector of dimension int(k/d) 1, where each baseline signal is of length K. iii iv Obtain the new signal at a later time, say t. This new signal could be obtained from a potentially damaged location. Compute the DB4 wavelet coefficients of the new signals at the first, fourth, fifth and sixth dyadic scales. Energies of these coefficients are calculated in a similar fashion as performed in step ii. These are represented by E [ E E E ] { 1 Q} new = 1, new, 5, new, 6, new, E7, new,,..., (2) where E i,new, is the energy of the wavelet coefficients of the th chunk of the new signal at the ith dyadic scale. 2 The data analyzed in this paper were stationary ambient vibration signals from the ASCE Benchmark Structure. The window size of 20 data points were chosen so that the probabilistic characteristics of the signals are preserved. In practice, the window size depends on sampling frequency and the correlation structure of the signal under question. 5

6 v Selection of closest baseline signal step: Search for the best baseline signal (obtained in Step i) with similar environmental and loading conditions. This is performed as follows: For each baseline signal, form a matrix Z 0, = [E 1,baseline, E 1,new, ] ( = 1,,Q) of dimension int(k/d) 2. Perform a principal components (PC) analysis on Z 0,. Obtain the principal direction, v 1 ( = 1,,Q). Also, obtain the minimum variance (square of the lowest singular value) and denote it as κ 0. This procedure is presented in Section 3. Choose the signal closest to the new signal as that signal with the value of τ = v v 11, 21, 1 and the lowest value of κ 0, where v 11, and v 21, are the values of the components of vector v 1. vi Feature extraction step: Extract features by comparing the energies of the new signal with that of the best signal. This is performed as follows: Form a matrix Z,k = [E i,best, E i,new, ] (i = 5,6,7; = 1,,Q; k = 1,2,3), where E i,best, is the energy of the wavelet coefficients of the th chunk of the best signal in the database at the ith dyadic scale. Perform a principal component analysis on Z,k ( = 1,,Q; k = 1,2,3), as explained in Section 3. Obtain the minimum variance (square of the lowest singular value) and denote it as κ,k ( = 1,,Q; k = 1,2,3). Choose the damage sensitive feature vector as κ = [κ 1 κ 2 κ 3 ] where κ 1 is the vector whose th component is κ,1. vii Classification step: Use the k-means clustering algorithm with the gap statistic to discriminate between a damaged state and an undamaged state. 6

7 Fix the number of clusters as k. For a fixed value of k, obtain the cluster centers of the grouped feature vectors X defined in Equation (3), using the k-means algorithm. κ undamaged X = (3) κ new where κ undamaged and κ new are the feature vectors obtained from an undamaged (the best/closest signal in the database to the new signal) and new signal respectively. Calculate the gap statistic (Tibshirani et al., 2001), as explained in Section 4. Use the gap statistic to determine the optimal number of clusters. If the number of clusters is greater than one, then it is hypothesized that some degree of damage has taken place. If, however, the clusters are very close based on the gap statistic, then it is concluded that there is no damage. Such signals would be stored in the baseline database. viii ix If there is damage, calculate the extent of damage by using the Euclidean distance between the means of the damaged and undamaged clusters. Go to step iii. 3. APPLICATION OF PRINCIPAL COMPONENTS ANALYSIS IN OPTIMAL SELECTION OF BASELINE SIGNAL AND FEATURE EXTRACTION Principal components analysis is used in the optimal baseline selection and in the feature extraction steps of the algorithm. In this section, the theory behind principal components is explained. Principal Components Analysis Principal components analysis (PCA) is a linear transformation used in multivariate statistical analysis in order to reduce the dimension of the dataset and to find patterns (or feature extraction) in the dataset [9]. The mathematical principle behind PCA is 7

8 explained as follows: Consider a centered matrix Y of dimension, say N p. The sample covariance matrix is given by S = Y T Y/N-1. Then the eigen decomposition of Y T Y is given as T 2 Y Y = VD V V = D = [ v1, v 2,..., v p ] diag [ d, d,..., d ] 1 T 2 p (4) where, V is a set of eigenvectors v i (i =1,..,p) and D is the singular value matrix. The eigenvectors v are called the principal component directions of Y. The first principal component direction v 1 has the property that z 1 = Yv 1 has the largest sample variance amongst all the linear combinations of the columns of Y [9]. Also, z 1 is called the first principal component. In this study, principal components (PC s) are not used as a device to reduce dimensionality but for feature extraction. This is explained in more detail in the sections below. Optimal Selection of Baseline Signal In practice, vibration data are collected under different operational conditions (loading amplitude and direction; and environmental conditions such as temperature and humidity). In order to compare ambient (or linear) vibration signals under various operational conditions, we will use the energies of the wavelet coefficients at the 1 st dyadic scale to determine whether there is any difference in the loading conditions of the signal or not. While comparing a signal (from a potentially damaged location) to a baseline signal, it is assumed that damage generally occurs in the lower frequency modes, which corresponds to higher scales. Thus, lower scales are chosen for comparing loading conditions on the assumption that damage will not affect the wavelet coefficients at these scales and thus similar wavelet energies at lower scales would effectively describe the loading conditions better. Moreover the wavelet coefficients at these scales will be able to take into account the transient phenomenon such as umps and spikes of the two signals being analyzed. 8

9 Figure 1 illustrates the plot of the energies of the wavelet coefficients at the first dyadic scale, for a similar and dissimilar vibration signal. The vector v 1 is plotted at 45 o to the x axis. In the case when we obtain acceleration datasets from similar loading conditions, it is noticed that the clouds cluster along the direction of v 1. The reason for this is that these values of E 1,baseline, and E 1,new, should be similar and thus cluster around v 1, implying a low variance in the direction of v 2. The best signal in the database closest to the new signal is selected using the following procedure: Form the matrix Z 0, = [E 1,baseline, E 1,new, ] ( = 1,,Q). Perform a principal component analysis on Z 0,. Z V D T 0, Z = = 0, = V D V 2 T [ v1 v2 ] { 1,..., Q} diag[ d, d ] 1 2 (4) where, V is a set of eigenvectors and D is the singular value matrix, obtained from the decomposition of Z 0,. Obtain the principal directions v 1 to calculate the ratio v 11, τ = and v21, 0, ( d ) 2 κ = (=1,,Q). Among all the signals in the database, find that signal in 2 the database with the values of the ratio τ approximately equal to one and having the lowest value of κ 0,. In the case when the signals are not similar, the value of E 1,new,1 and E 1,baseline,1 would be different and thus deviate away from the vector v 1. This is measured by τ v = v 11, 21,. If this quantity is closer to 1 (implying that the vector v 1 is at 45 o to the x-axis) it would mean that the energies of the signals are similar as opposed to large value of τ. In addition, the variance (spread) in the direction v 2 would give an indication of similarity. In Figure 1, a comparison of two dissimilar signals (Cluster 2) is illustrated to contrast two signals obtained from similar loading conditions (Cluster 1). To this end, the 9

10 principal directions will be used to obtain the directions of highest variances. The variance in the direction of the second principal direction is used to show how dissimilar the signals are. This is denoted as κ 0, whose th component is κ 0,. Thus, the lower the value of κ 0,, the more similar the signals are. The reason why the first scale is chosen is because the coefficients at the first scale will be able to detect transient phenomenon, which is a good descriptor of loading conditions. This assumes that damage does not affect the higher frequency modes of vibration. Figure 2 illustrates the comparison of two loading conditions as defined in the ASCE Benchmark Structure Experiment. The first excitation is a series of independent filtered Gaussian white noise loads generated using a sixth - order low-pass Butterworth filter with a 100 Hz cutoff and applied at each story of the structure. The second loading is a random excitation generated by a shaker on the roof-top of the center column. Figure 2 (a) shows the histogram of τ when comparing an acceleration signal from sensor 2 for damage pattern 2 to baseline signals recorded for the same loading conditions. The values of τ are in the range of , indicating that loading conditions are similar. Also it is noted that even though damage pattern 2 is the most severe of the damage patterns, it does not affect the values of τ. Similarly, Figure 2 (b) shows the histogram of τ when comparing an acceleration signal from sensor 2 for an undamaged signal with baseline signals recorded for different loading conditions. The values of τ are much higher than one and are in the range of , indicating that loading conditions are not similar. Given that there is a subset of signals in the database with value of τ close to 1 (Figure 2(a)), it is necessary to identify the optimal signal in this subset closest to the new signal. This is achieved by choosing the minimum value of κ 0. Figure 3 shows the variation of κ 0 for similar loading conditions while comparing a signal from sensor 2 and damage pattern 2 to a database of eighty similar baseline signals. The lowest value of κ 0 is chosen as the best signal in the database, closest to the new signal. It should be noted that values of κ 0 is in the range of to Thus choosing any of these eighty baseline 10

11 signals would suffice for our analysis. Since the goal of this study is in automating the damage detection algorithm, we choose the optimal signal to be the one with the minimum value of κ 0. Feature Extraction In previous papers by the authors [5, 6], closed form analytical solutions of wavelet energies have been derived for single degree and multiple degree of freedom systems, thus relating it to the parameters of the structural system such as natural frequencies, damping and mode shapes. The damage sensitive feature κ is obtained by performing the principal components analysis of the wavelet energies at the 5 th, 6 th and 7 th dyadic scales. Figure 4 illustrates the feature extraction procedure. Values of the energies of the wavelet coefficients at the fifth dyadic scale of the best baseline signal E 5,best and the new signal E 5,new are compared. Principal components analysis (PCA) is performed on this dataset and the variance in the second principal direction v 2, denoted as κ 1, is chosen as the first element of the feature vector κ. Similarly κ 2 and κ 3 (second and third elements of the feature vector) are derived using the energies at the sixth and seventh dyadic scales respectively. Intuitively, it can be understood that a large value of κ is a good indicator of damage. The damage sensitive features are extracted using the following procedure: Form a matrix Z,k = [E i,best, E i,new, ] (i = 5,6,7; = 1,,Q; k = 1,2,3), where E i,best, is the energy of the wavelet coefficients of the th chunk of the best signal in the database at the ith dyadic scale. Perform a principal component analysis on Z,k ( = 1,,Q; k = 1,2,3), as explained in Section 3. Obtain the minimum variance (square of the lowest singular value) and denote it as κ,k ( = 1,,Q; k = 1,2,3). Choose the damage sensitive feature vector as κ = [κ 1 κ 2 κ 3 ] where κ 1 is the vector whose th component is κ,1. It should be noted that the fifth, sixth and seventh dyadic scales are chosen by trial and error. Thus, when using this algorithm with a real structure, a finite element model of 11

12 the structure under question should be developed. Damage should then be induced on the finite element model and the most sensitive scales can be chosen from a similar analysis as stated in this section. In a recent study by Noh and Kiremidian [8], the authors have found the optimal scale for feature extraction to be closely connected to the instantaneous frequency, which corresponds to that pseudo frequency at which wavelet coefficients attains a maximum value at a particular time t. In other words, the instantaneous frequency is the dominant frequency at time t. It is also shown that the instantaneous frequency is closely related to the natural frequency of the structure [8]. Figure 5 illustrates the variation of the damage sensitive feature for an undamaged case, maor damage pattern 1 and minor damage pattern 6. The ASCE Benchmark experiment is described in the Appendix and in more detail in [7]. It is noted that the values of the damage sensitive feature for an undamaged case are close to zero. However, for damage patterns 1 and 6, the values of the feature vector are significantly greater than zero and form separate clouds with respect to the undamaged cloud, thus indicating damage. From Figure 5, it is also observed that as the level of damage increases the distance of the cloud from the origin increases. In the next section, a classification methodology to automate the damage detection process using the k-means algorithm and the gap statistic is explained. Also, a damage measure DM using the Euclidean distance between the means of the undamaged and damaged feature vectors is developed. 4. DAMAGE DIAGNOSIS The main premise in the developed algorithm is that there is a migration of the feature vector with the onset of damage. To automate the damage decision process, the k-means algorithm and the gap statistic are used. The k-means algorithm is a commonly used hard clustering scheme [10]. In this particular study, the k-means algorithm is used instead of Gaussian mixtures modeling [11], since there is a larger separation of clouds while using wavelet based feature vectors (Figure 5) as in comparison to the same study with the AR coefficients as feature vectors [11]. A damage measure DM using the Euclidean distance between the means of the undamaged and damaged feature vectors is also presented in this section. 12

13 Damage Detection using the k-means Algorithm and the Gap Statistic Figure 5 shows the results from the application of the proposed damage algorithm to the numerically simulated datasets of the ASCE Benchmark structure. From Figure 5, it can be observed that there is a distinct separation in the clouds of damage sensitive feature vector κ for damage patterns 1 and 6 with the respect to the undamaged case. The k-means algorithm is used to obtain the cluster centers with the prior assumption of the number of clusters. The gap statistic is computed to obtain the optimal number of cluster in the dataset [12]. The computation of the gap statistic is described below [12]. Fix the number of clusters and then use the k-means algorithm to cluster the observed data. For each of these cases, calculate W k : k = 1,.., M, where W k is a dispersion measure based on the within cluster distance defined in Tibshirani et al., 2001 [11]. Generate B reference datasets according to the uniform distribution and calculate the dispersion measure W kb for all b = 1,,B, and k=1,,m. Compute the mean and the standard deviation as follows µ sd s c ( W ) ( k ) = ( log( W ) µ ) ( k ) 1 = B b = sd 1 B log b 1 1+ B kb kb c (6) Choose the number of clusters by using the rule given below kˆ = smallest k such that Gap Gap 1 B B ( k ) ( k ) = log( W ) log( W ) b= 1 kb Gap( k + 1) s k k + 1 (7) The damage detection using the k-means algorithm and the gap statistic is performed as follows: For k = 1 to M (M is the number of clusters) 13

14 o Initialize the means of the dataset to k randomly chosen points o For each cluster mean, µ, (= 1,,k), find the points in the dataset closest to µ. Denote these set of points as C and the number of points as n. o Compute the new mean 3 μ = 1 n x i C x i (8) where, x i is a vector belonging to cluster, C. o o Iterate the above two steps until convergence is obtained. After computing the means of the dataset, the gap statistic is computed. Using the rules in Equation 7, find the optimal number of clusters in the dataset. From the application of this algorithm, if the number of clusters k, is found to be greater than one, it is hypothesized that the signals come from different states of the structure and this change of state is most likely due to damage. Figure 6 illustrates the migration of clusters for an undamaged feature vector and damage pattern (DP) 6. It is observed that even though DP 6 is a minor damage pattern, there is a large separation between the feature vectors before and after damage. The k- means algorithm is used to obtain the cluster centers in the dataset. The gap statistic predicts that there are 2 clusters in the dataset, indicating that there is damage. Figure 7 and Figure 8 illustrate a similar trend in the migration of the feature vectors for damage patterns 1-4, as defined in the ASCE Benchmark Experiment. In all these cases, the gap statistic predicts that there are two clusters in the dataset, thus 3 The dataset used for damage detection is given by the grouped feature vector X as defined in Equation 3. 14

15 indicating damage. The results for all sensors in the ASCE Benchmark Experiment are given in the next subsection, where the damage extent measure is developed. Damage Extent Measure The damage sensitive feature is defined as follows: ˆ T ( ˆ μˆ ) ( μˆ ) DM = μ μ (9), ˆ ˆ κ, new κ, undamaged κ, new κ undamaged where, µ κ, new and µ κ, undamaged are the sample means of κ new and κ undamaged respectively. Table 1 shows the variation of DM for all sensors and damage patterns as defined in the ASCE Benchmark Experiment. From Table 1, the following observations are made The values of DM are correlated to the amount of damage. For damage patterns 1 and 2, the values of DM are similar. A similar observation is made for damage patterns 3 and 4 for sensors placed on Faces 2 and 4 of the structure (Figure 9). For minor damage patterns 3 and 6, the values of DM are high for sensors on Faces 2 and 4 in comparison to Faces 1 and 3. It is also observed that no damage is detected at sensors on Faces 1 and 3 for damage pattern 6. These are denoted as NA in Table 1. This behavior is expected since damage patterns 3 and 6 involves the reduction of stiffness of a brace on Face 2 of the ASCE Benchmark Structure (Figure 9). For sensors instrumented on Faces 1 and 3 of the ASCE Benchmark Structure (Figure 9), the values of DM are significantly higher for damage pattern 4 in comparison to damage pattern 3. For example, in the case of sensor 1, the value of DM increases from for DP 3 to for DP 4. The reason for this trend is because DP 3 involves the removal of a brace on Face 2, whereas DP 4 is achieved by removing a brace on Face 1 and Face 2 of the structure. 15

16 All damage patterns are detected consistently using the wavelet based feature vector. For all damage patterns, the values of DM are higher for sensors on Faces 2 and 4. The reason for this behavior may be attributed to the fact that this is the weak direction of the structure. CONCLUSIONS This paper describes the development of a wavelet based damage diagnosis algorithm using principal components analysis. This algorithm requires the creation of a database of baseline measurements (which can include accelerations and strains). The first part of the paper describes a data preprocessing algorithm which enables the estimation of the optimal baseline signal closest to the new signal. Once the optimal baseline signal in the database is chosen, the damage sensitive feature vector is calculated as a function of the wavelet energies at the fifth, sixth and seventh dyadic scales. The process of damage detection is automated using the k-means algorithm and the Gap statistic. A simple damage measure based on the Euclidean distance between the means of the damaged and undamaged datasets is also described. The k-means algorithm and Gap statistic works consistently well for detecting damage patterns as defined for the ASCE Benchmark Structure. The values of DM also correlate well with the extent of damage. Although these results are promising, practical implementation of such an algorithm would require further testing and evaluation with experimental and field data. This would involve collecting structural response data under varying degrees of damage, loading conditions, environmental conditions such as temperature and humidity, as well as different sequences of damage occurrences. This would result in the creation of a larger database of baseline signals thus enabling the testing of the pre-processing, damage detection and quantification aspects of the developed algorithm. In addition, wireless sensors with this damage detection algorithm embedded on a microprocessor have to be developed and tested on experimental and real structures. It is only after extensive experimentation and field testing with calibration that these models can be widely applied. Never-the-less, the results presented here are 16

17 encouraging and represent a good initial step towards achieving this goal. Current research is being focused on the reliability of these algorithms for strong motion earthquake data [8, 13]. APPENDIX The ASCE Benchmark experiment is described in this appendix. The ASCE Benchmark structure is a four storey, 2 bay by 2 bay structure. The placement of the sensors and the direction of the acceleration measurements are shown in Figure 9. Damage is simulated by removal of braces thus resulting in the loss of storey stiffness and also by bolt loosening. The damage patterns include: Damage pattern 1: Removal of all braces on the first floor Damage pattern 2: Removal of all braces on the first and third floors Damage pattern 3: Removal of one brace (near sensor 2 measuring acceleration in the y-direction) on the first floor Damage pattern 4: Removal of one brace on the first (near sensor 2 measuring acceleration in the y-direction) and third (near sensor 9 measuring acceleration in the x-direction) floors Damage pattern 5: Damage Pattern 4 + loosening of bolts Damage pattern 6: Reduction of the stiffness of a brace to 1/3 rd of its original References value (near sensor 2 measuring acceleration in the y-direction) 1. Straser, E. G. and Kiremidian, A. S. (1998). Modular wireless damage monitoring system for structures., Report No. 128, John A. Blume Earthquake Engineering Center, Department of Civil and Environmental Engineering, Stanford University, Stanford, CA. 2. Lynch, J. P., Sundararaan, A., Law, K. H., Kiremidian, A.S. and Carryer, E. (2004). Embedding damage detection algorithms in a wireless sensing unit for attainment of operational power efficiency. Smart Materials and Structures, 13(4), Sohn, H., Farrar, C. R., Hunter, H. F., and Worden, K. (2001). Applying the LANL statistical pattern recognition paradigm for structural health monitoring to data from a surface-effect fast patrol boat, LANL Report LA MS, LANL Los Alamos, NM

18 4. Nair, K. K., Kiremidian, A. S. and Law, K. H. (2006). Time series based damage detection and localization algorithm with application to the ASCE benchmark structure. Journal of Sound and Vibration, 291 (2), Nair, K. K. and Kiremidian, A. S. (2009). Damage Detection Using Haar Wavelet Transforms of Vibration Signals. Forthcoming in ASME Journal of Applied Mechanics. 6. Nair, K. K. and Kiremidian, A. S. (2009). A Morlet Wavelet Based Damage Detection Algorithm. Submitted. 7. Johnson, E. A., Lam, H. F., Katafygiotis, L. S. and Beck, J. L. (2004). Phase I IASC- ASCE structural health monitoring benchmark problem using simulated data. ASCE Journal of Engineering Mechanics, 130(1), Noh, H. and Kiremidian, A. S. (2009). On the use of wavelet coefficient energy for structural damage diagnosis. 10 th International Conference on Structural Safety and Reliability, Osaka, Japan. 9. Mardia, K. V., Kent, J. T. and Bibby, J. M. (2003). Multivariate Analysis, Academic Press, London. 10. Hastie, T., Tibshirani, R. and Freidman, J., (2001). Elements of Statistical Learning: Data Mining, Inference and Prediction, Springer Verlag, First edition, New York. 11. Nair, K. K. and Kiremidian, A. S. (2007). Time series based structural damage detection algorithm using Gaussian mixtures modeling. ASME Journal of Dynamic Systems, Measurement and Control, 129(3), Tibshirani, R., Walther, G. and Hastie, T., Estimating the Number of Clusters in a Dataset via the Gap Statistic, J. of the Royal Stat. Soc: Series B, 63(2), pp Noh, H., Nair, K. K, Lignos, D. and Kiremidian, A. S. On the use of wavelet based damage sensitive features for structural damage diagnosis, Under preparation. 18

19 Figures Cluster 2 v 1 E 1, new,1 Similar signal v 2 Cluster 1 v 2 Dissimilar signal v 1 E 1, baseline,1 Figure 1: Illustration of a similar and dissimilar cloud by comparing E 1,baseline and E 1,new 19

20 (a) (b) Figure 2: Histogram of τ for sensor 2 for (a) similar loading condition with DP2 and (b) dissimilar loading conditions for undamaged cases 20

21 κ0 Number of records Figure 3: Variation of κ 0 for similar loading conditions comparing undamaged case and damage pattern 2 21

22 Damaged cloud v 2 v 1 E 5, damaged κ 1 = Variance in direction v 2 v 4 v 3 Undamaged cloud E 5, best Figure 4: Illustration of damaged and undamaged cloud using principal components analysis 22

23 Figure 5: Variation of the damage sensitive feature vectors for damage patterns (DP) 0, 1 and 6 as defined in the ASCE Benchmark Experiment 23

24 (a) (b) Figure 6: Migration of the feature vectors κ with damage for minor patterns (a) Damage pattern 6 and (b) a zoom in of the undamaged cloud (Undamaged ο; Damaged +) 24

25 (a) (b) Figure 7: Migration of the feature vectors with damage for damage patterns (a) Damage pattern 3 and (b) Damage Pattern 4 (Undamaged ο; Damaged +) 25

26 (a) (b) Figure 8: Migration of the feature vectors with damage for maor patterns (a) Damage pattern 1 and (b) Damage Pattern 2 (Undamaged ο; Damaged +) 26

27 Figure 9: Placement of sensors and directions of acceleration measurements on the ASCE Benchmark Structure [7] 27

28 Tables Table 1: Variation of DM for the Morlet wavelet based damage sensitive feature for various sensors and different damage patterns Sensor DP1 DP2 DP3 DP4 DP5 DP NA NA NA NA NA NA NA NA

Statistical Damage Detection Using Time Series Analysis on a Structural Health Monitoring Benchmark Problem

Statistical Damage Detection Using Time Series Analysis on a Structural Health Monitoring Benchmark Problem Source: Proceedings of the 9th International Conference on Applications of Statistics and Probability in Civil Engineering, San Francisco, CA, USA, July 6-9, 2003. Statistical Damage Detection Using Time

More information

On the use of wavelet coefficient energy for structural damage diagnosis

On the use of wavelet coefficient energy for structural damage diagnosis Safety, Reliability and Risk of Structures, Infrastructures and Engineering Systems Furuta, Frangopol & Shinozuka (eds) 010 Taylor & Francis Group, London, ISBN 978-0-415-47557-0 On the use of wavelet

More information

Damage Characterization of the IASC-ASCE Structural Health Monitoring Benchmark Structure by Transfer Function Pole Migration. J. P.

Damage Characterization of the IASC-ASCE Structural Health Monitoring Benchmark Structure by Transfer Function Pole Migration. J. P. SOURCE: Jerome P. Lynch, " Damage Characterization of the IASC-ASCE Structural Health Monitoring Benchmark Structure by Transfer Function Pole Migration, " Proceedings of the 2005 ASCE Structures Congress,

More information

The application of statistical pattern recognition methods for damage detection to field data

The application of statistical pattern recognition methods for damage detection to field data IOP PUBLISHING Smart Mater. Struct. 17 (2008) 065023 (12pp) SMART MATERIALS AND STRUCTURES doi:10.1088/0964-1726/17/6/065023 The application of statistical pattern recognition methods for damage detection

More information

Time series-based damage detection and localization algorithm with application to the ASCE benchmark structure

Time series-based damage detection and localization algorithm with application to the ASCE benchmark structure Journal of Sound and Vibration 291 (2006) 349 368 JOURNAL OF SOUND AND VIBRATION www.elsevier.com/locate/jsvi Time series-based damage detection and localization algorithm with application to the ASCE

More information

Ph.D student in Structural Engineering, Department of Civil Engineering, Ferdowsi University of Mashhad, Azadi Square, , Mashhad, Iran

Ph.D student in Structural Engineering, Department of Civil Engineering, Ferdowsi University of Mashhad, Azadi Square, , Mashhad, Iran Alireza Entezami a, Hashem Shariatmadar b* a Ph.D student in Structural Engineering, Department of Civil Engineering, Ferdowsi University of Mashhad, Azadi Square, 9177948974, Mashhad, Iran b Associate

More information

EMBEDDING ALGORITHMS IN A WIRELESS STRUCTURAL MONITORING SYSTEM

EMBEDDING ALGORITHMS IN A WIRELESS STRUCTURAL MONITORING SYSTEM Source: Proceedings of International Conference on Advances and New Challenges in Earthquae Engineering Research (ICANCEER02), Hong Kong, China, August 9-20, 2002. EMBEDDING ALGORITHMS IN A WIRELESS STRUCTURAL

More information

Hierarchical sparse Bayesian learning for structural health monitoring. with incomplete modal data

Hierarchical sparse Bayesian learning for structural health monitoring. with incomplete modal data Hierarchical sparse Bayesian learning for structural health monitoring with incomplete modal data Yong Huang and James L. Beck* Division of Engineering and Applied Science, California Institute of Technology,

More information

stiffness to the system stiffness matrix. The nondimensional parameter i is introduced to allow the modeling of damage in the ith substructure. A subs

stiffness to the system stiffness matrix. The nondimensional parameter i is introduced to allow the modeling of damage in the ith substructure. A subs A BAYESIAN PROBABILISTIC DAMAGE DETECTION USING LOAD-DEPENDENT RIT VECTORS HOON SOHN Λ and KINCHO H. LAW y Department of Civil and Environmental Engineering, Stanford University, Stanford, CA 9435-42,U.S.A.

More information

ABSTRACT INTRODUCTION

ABSTRACT INTRODUCTION ABSTRACT Presented in this paper is an approach to fault diagnosis based on a unifying review of linear Gaussian models. The unifying review draws together different algorithms such as PCA, factor analysis,

More information

Bayesian System Identification based on Hierarchical Sparse Bayesian Learning and Gibbs Sampling with Application to Structural Damage Assessment

Bayesian System Identification based on Hierarchical Sparse Bayesian Learning and Gibbs Sampling with Application to Structural Damage Assessment Bayesian System Identification based on Hierarchical Sparse Bayesian Learning and Gibbs Sampling with Application to Structural Damage Assessment Yong Huang a,b, James L. Beck b,* and Hui Li a a Key Lab

More information

Reliable Condition Assessment of Structures Using Uncertain or Limited Field Modal Data

Reliable Condition Assessment of Structures Using Uncertain or Limited Field Modal Data Reliable Condition Assessment of Structures Using Uncertain or Limited Field Modal Data Mojtaba Dirbaz Mehdi Modares Jamshid Mohammadi 6 th International Workshop on Reliable Engineering Computing 1 Motivation

More information

Unsupervised Learning Methods

Unsupervised Learning Methods Structural Health Monitoring Using Statistical Pattern Recognition Unsupervised Learning Methods Keith Worden and Graeme Manson Presented by Keith Worden The Structural Health Monitoring Process 1. Operational

More information

SENSITIVITY ANALYSIS OF ADAPTIVE MAGNITUDE SPECTRUM ALGORITHM IDENTIFIED MODAL FREQUENCIES OF REINFORCED CONCRETE FRAME STRUCTURES

SENSITIVITY ANALYSIS OF ADAPTIVE MAGNITUDE SPECTRUM ALGORITHM IDENTIFIED MODAL FREQUENCIES OF REINFORCED CONCRETE FRAME STRUCTURES SENSITIVITY ANALYSIS OF ADAPTIVE MAGNITUDE SPECTRUM ALGORITHM IDENTIFIED MODAL FREQUENCIES OF REINFORCED CONCRETE FRAME STRUCTURES K. C. G. Ong*, National University of Singapore, Singapore M. Maalej,

More information

Machine Learning 11. week

Machine Learning 11. week Machine Learning 11. week Feature Extraction-Selection Dimension reduction PCA LDA 1 Feature Extraction Any problem can be solved by machine learning methods in case of that the system must be appropriately

More information

Sequential Damage Detection based on the Continuous Wavelet Transform

Sequential Damage Detection based on the Continuous Wavelet Transform Sequential Damage Detection based on the Continuous Wavelet Transform Yizheng Liao a and Konstantinos Balafas b and Ram Rajagopal a and Anne S. Kiremidjian b a Stanford Sustainable Systems Lab, Department

More information

WILEY STRUCTURAL HEALTH MONITORING A MACHINE LEARNING PERSPECTIVE. Charles R. Farrar. University of Sheffield, UK. Keith Worden

WILEY STRUCTURAL HEALTH MONITORING A MACHINE LEARNING PERSPECTIVE. Charles R. Farrar. University of Sheffield, UK. Keith Worden STRUCTURAL HEALTH MONITORING A MACHINE LEARNING PERSPECTIVE Charles R. Farrar Los Alamos National Laboratory, USA Keith Worden University of Sheffield, UK WILEY A John Wiley & Sons, Ltd., Publication Preface

More information

SYSTEM IDENTIFICATION & DAMAGE ASSESSMENT OF STRUCTURES USING OPTICAL TRACKER ARRAY DATA

SYSTEM IDENTIFICATION & DAMAGE ASSESSMENT OF STRUCTURES USING OPTICAL TRACKER ARRAY DATA SYSTEM IDENTIFICATION & DAMAGE ASSESSMENT OF STRUCTURES USING OPTICAL TRACKER ARRAY DATA Chin-Hsiung Loh 1,* and Chuan-Kai Chan 1 1 Department of Civil Engineering, National Taiwan University Taipei 10617,

More information

The effect of environmental and operational variabilities on damage detection in wind turbine blades

The effect of environmental and operational variabilities on damage detection in wind turbine blades The effect of environmental and operational variabilities on damage detection in wind turbine blades More info about this article: http://www.ndt.net/?id=23273 Thomas Bull 1, Martin D. Ulriksen 1 and Dmitri

More information

Vibration Based Health Monitoring for a Thin Aluminum Plate: Experimental Assessment of Several Statistical Time Series Methods

Vibration Based Health Monitoring for a Thin Aluminum Plate: Experimental Assessment of Several Statistical Time Series Methods Vibration Based Health Monitoring for a Thin Aluminum Plate: Experimental Assessment of Several Statistical Time Series Methods Fotis P. Kopsaftopoulos and Spilios D. Fassois Stochastic Mechanical Systems

More information

Local Damage Detection in Beam-Column Connections Using a Dense Sensor Network

Local Damage Detection in Beam-Column Connections Using a Dense Sensor Network 3143 Local Damage Detection in Beam-Column Connections Using a Dense Sensor Network Elizabeth L. Labuz 1, Minwoo Chang 2, and Shamim N. Pakzad 3 1 Graduate Student, Department of Civil and Environmental

More information

A Probabilistic Framework for solving Inverse Problems. Lambros S. Katafygiotis, Ph.D.

A Probabilistic Framework for solving Inverse Problems. Lambros S. Katafygiotis, Ph.D. A Probabilistic Framework for solving Inverse Problems Lambros S. Katafygiotis, Ph.D. OUTLINE Introduction to basic concepts of Bayesian Statistics Inverse Problems in Civil Engineering Probabilistic Model

More information

ECE 661: Homework 10 Fall 2014

ECE 661: Homework 10 Fall 2014 ECE 661: Homework 10 Fall 2014 This homework consists of the following two parts: (1) Face recognition with PCA and LDA for dimensionality reduction and the nearest-neighborhood rule for classification;

More information

Introduction to Statistical Inference

Introduction to Statistical Inference Structural Health Monitoring Using Statistical Pattern Recognition Introduction to Statistical Inference Presented by Charles R. Farrar, Ph.D., P.E. Outline Introduce statistical decision making for Structural

More information

Dynamic System Identification using HDMR-Bayesian Technique

Dynamic System Identification using HDMR-Bayesian Technique Dynamic System Identification using HDMR-Bayesian Technique *Shereena O A 1) and Dr. B N Rao 2) 1), 2) Department of Civil Engineering, IIT Madras, Chennai 600036, Tamil Nadu, India 1) ce14d020@smail.iitm.ac.in

More information

A methodology for fault detection in rolling element bearings using singular spectrum analysis

A methodology for fault detection in rolling element bearings using singular spectrum analysis A methodology for fault detection in rolling element bearings using singular spectrum analysis Hussein Al Bugharbee,1, and Irina Trendafilova 2 1 Department of Mechanical engineering, the University of

More information

Post-earthquake Damage Detection Using Embedded Electro-mechanical Impedance Sensors for Concrete Dams

Post-earthquake Damage Detection Using Embedded Electro-mechanical Impedance Sensors for Concrete Dams Post-earthquake Damage Detection Using Embedded Electro-mechanical Impedance Sensors for Concrete Dams X. Feng, E.T. Dandjekpo & J. Zhou Faculty of Infrastructure, Dalian University of Technology, China

More information

EXPERIMENTAL EVALUATION OF MODAL PARAMETER VARIATIONS FOR STRUCTURAL HEALTH MONITORING

EXPERIMENTAL EVALUATION OF MODAL PARAMETER VARIATIONS FOR STRUCTURAL HEALTH MONITORING EXPERIMENTAL EVALUATION OF MODAL PARAMETER VARIATIONS FOR STRUCTURAL HEALTH MONITORING Ruben L. Boroschek 1, Patricio A. Lazcano 2 and Lenart Gonzalez 3 ABSTRACT : 1 Associate Professor, Dept. of Civil

More information

Application of load-dependent Ritz vectors to Bayesian probabilistic damage detection

Application of load-dependent Ritz vectors to Bayesian probabilistic damage detection Probabilistic Engineering Mechanics 15 (2000) 139 153 www.elsevier.com/locate/probengmech Application of load-dependent Ritz vectors to Bayesian probabilistic damage detection Hoon Sohn, Kincho H. Law*

More information

VARIANCE COMPUTATION OF MODAL PARAMETER ES- TIMATES FROM UPC SUBSPACE IDENTIFICATION

VARIANCE COMPUTATION OF MODAL PARAMETER ES- TIMATES FROM UPC SUBSPACE IDENTIFICATION VARIANCE COMPUTATION OF MODAL PARAMETER ES- TIMATES FROM UPC SUBSPACE IDENTIFICATION Michael Döhler 1, Palle Andersen 2, Laurent Mevel 1 1 Inria/IFSTTAR, I4S, Rennes, France, {michaeldoehler, laurentmevel}@inriafr

More information

Operational modal analysis using forced excitation and input-output autoregressive coefficients

Operational modal analysis using forced excitation and input-output autoregressive coefficients Operational modal analysis using forced excitation and input-output autoregressive coefficients *Kyeong-Taek Park 1) and Marco Torbol 2) 1), 2) School of Urban and Environment Engineering, UNIST, Ulsan,

More information

Application of Extreme Value Statistics for Structural Health Monitoring. Hoon Sohn

Application of Extreme Value Statistics for Structural Health Monitoring. Hoon Sohn Application of Etreme Value Statistics for Structural Health Monitoring Hoon Sohn Weapon Response Group Engineering Sciences and Applications Division Los Alamos National Laboratory Los Alamos, New Meico,

More information

EVALUATION OF THE ENVIRONMENTAL EFFECTS ON A MEDIUM RISE BUILDING

EVALUATION OF THE ENVIRONMENTAL EFFECTS ON A MEDIUM RISE BUILDING 7th European Workshop on Structural Health Monitoring July 8-11, 214. La Cité, Nantes, France More Info at Open Access Database www.ndt.net/?id=17121 EVALUATION OF THE ENVIRONMENTAL EFFECTS ON A MEDIUM

More information

Switch Mechanism Diagnosis using a Pattern Recognition Approach

Switch Mechanism Diagnosis using a Pattern Recognition Approach The 4th IET International Conference on Railway Condition Monitoring RCM 2008 Switch Mechanism Diagnosis using a Pattern Recognition Approach F. Chamroukhi, A. Samé, P. Aknin The French National Institute

More information

Using SDM to Train Neural Networks for Solving Modal Sensitivity Problems

Using SDM to Train Neural Networks for Solving Modal Sensitivity Problems Using SDM to Train Neural Networks for Solving Modal Sensitivity Problems Brian J. Schwarz, Patrick L. McHargue, & Mark H. Richardson Vibrant Technology, Inc. 18141 Main Street Jamestown, California 95327

More information

Damage detection of truss bridge via vibration data using TPC technique

Damage detection of truss bridge via vibration data using TPC technique Damage detection of truss bridge via vibration data using TPC technique Ahmed Noor AL-QAYYIM 1,2, Barlas Özden ÇAĞLAYAN 1 1 Faculty of Civil Engineering, Istanbul Technical University, Istanbul, Turkey

More information

Transmissibility Function Analysis for Boundary Damage Identification of a Two-Storey Framed Structure using Artificial Neural Networks

Transmissibility Function Analysis for Boundary Damage Identification of a Two-Storey Framed Structure using Artificial Neural Networks Transmissibility Function Analysis for Boundary Damage Identification of a Two-Storey Framed Structure using Artificial Neural Networks U. Dackermann, J. Li & B. Samali Centre for Built Infrastructure

More information

Comparison of the Results Inferred from OMA and IEMA

Comparison of the Results Inferred from OMA and IEMA Comparison of the Results Inferred from OMA and IEMA Kemal Beyen, Kocaeli University, Kocaeli, Turkey, kbeyen@kocaeli.edu.tr Mustafa Kutanis Sakarya University, Sakarya, Turkey, mkutanis@gmail.com.tr ABSTRACT:

More information

Forecasting Wind Ramps

Forecasting Wind Ramps Forecasting Wind Ramps Erin Summers and Anand Subramanian Jan 5, 20 Introduction The recent increase in the number of wind power producers has necessitated changes in the methods power system operators

More information

Principal Component Analysis vs. Independent Component Analysis for Damage Detection

Principal Component Analysis vs. Independent Component Analysis for Damage Detection 6th European Workshop on Structural Health Monitoring - Fr..D.4 Principal Component Analysis vs. Independent Component Analysis for Damage Detection D. A. TIBADUIZA, L. E. MUJICA, M. ANAYA, J. RODELLAR

More information

Motivating the Covariance Matrix

Motivating the Covariance Matrix Motivating the Covariance Matrix Raúl Rojas Computer Science Department Freie Universität Berlin January 2009 Abstract This note reviews some interesting properties of the covariance matrix and its role

More information

OBSERVER/KALMAN AND SUBSPACE IDENTIFICATION OF THE UBC BENCHMARK STRUCTURAL MODEL

OBSERVER/KALMAN AND SUBSPACE IDENTIFICATION OF THE UBC BENCHMARK STRUCTURAL MODEL OBSERVER/KALMAN AND SUBSPACE IDENTIFICATION OF THE UBC BENCHMARK STRUCTURAL MODEL Dionisio Bernal, Burcu Gunes Associate Proessor, Graduate Student Department o Civil and Environmental Engineering, 7 Snell

More information

2330. A study on Gabor frame for estimating instantaneous dynamic characteristics of structures Wei-Chih Su 1, Chiung-Shiann Huang 2 1

2330. A study on Gabor frame for estimating instantaneous dynamic characteristics of structures Wei-Chih Su 1, Chiung-Shiann Huang 2 1 2330. A study on Gabor frame for estimating instantaneous dynamic characteristics of structures Wei-Chih Su 1, Chiung-Shiann Huang 2 1 National Center for High-Performance Computing, National Applied Research

More information

DAMAGE DETECTION WITH INTERVAL ANALYSIS FOR UNCERTAINTIES QUANTIFICATION

DAMAGE DETECTION WITH INTERVAL ANALYSIS FOR UNCERTAINTIES QUANTIFICATION DAMAGE DETECTION WITH INTERVAL ANALYSIS FOR UNCERTAINTIES QUANTIFICATION Gang Liu 1, 2,*, Zhu Mao 3, Jun Luo 2 1. Key Laboratory of New Technology for Construction of Cities in Mountain Area (Chongqing

More information

Impeller Fault Detection for a Centrifugal Pump Using Principal Component Analysis of Time Domain Vibration Features

Impeller Fault Detection for a Centrifugal Pump Using Principal Component Analysis of Time Domain Vibration Features Impeller Fault Detection for a Centrifugal Pump Using Principal Component Analysis of Time Domain Vibration Features Berli Kamiel 1,2, Gareth Forbes 2, Rodney Entwistle 2, Ilyas Mazhar 2 and Ian Howard

More information

Intelligent Fault Classification of Rolling Bearing at Variable Speed Based on Reconstructed Phase Space

Intelligent Fault Classification of Rolling Bearing at Variable Speed Based on Reconstructed Phase Space Journal of Robotics, Networking and Artificial Life, Vol., No. (June 24), 97-2 Intelligent Fault Classification of Rolling Bearing at Variable Speed Based on Reconstructed Phase Space Weigang Wen School

More information

Stationary or Non-Stationary Random Excitation for Vibration-Based Structural Damage Detection? An exploratory study

Stationary or Non-Stationary Random Excitation for Vibration-Based Structural Damage Detection? An exploratory study 6th International Symposium on NDT in Aerospace, 12-14th November 2014, Madrid, Spain - www.ndt.net/app.aerondt2014 More Info at Open Access Database www.ndt.net/?id=16938 Stationary or Non-Stationary

More information

Feature selection and classifier performance in computer-aided diagnosis: The effect of finite sample size

Feature selection and classifier performance in computer-aided diagnosis: The effect of finite sample size Feature selection and classifier performance in computer-aided diagnosis: The effect of finite sample size Berkman Sahiner, a) Heang-Ping Chan, Nicholas Petrick, Robert F. Wagner, b) and Lubomir Hadjiiski

More information

STRUCTURAL DAMAGE DETECTION USING FREQUENCY RESPONSE FUNCTIONS

STRUCTURAL DAMAGE DETECTION USING FREQUENCY RESPONSE FUNCTIONS STRUCTURAL DAMAGE DETECTION USING FREQUENCY RESPONSE FUNCTIONS A Thesis by SELCUK DINCAL Submitted to the Office of Graduate Studies of Texas A&M University in partial fulfillment of the requirements for

More information

Analyzing the Effect of Moving Resonance on Seismic Response of Structures Using Wavelet Transforms

Analyzing the Effect of Moving Resonance on Seismic Response of Structures Using Wavelet Transforms Analyzing the Effect of Moving Resonance on Seismic Response of Structures Using Wavelet Transforms M.R. Eatherton Virginia Tech P. Naga WSP Cantor Seinuk, New York, NY SUMMARY: When the dominant natural

More information

Relevance Vector Machines for Earthquake Response Spectra

Relevance Vector Machines for Earthquake Response Spectra 2012 2011 American American Transactions Transactions on on Engineering Engineering & Applied Applied Sciences Sciences. American Transactions on Engineering & Applied Sciences http://tuengr.com/ateas

More information

Machine Learning. B. Unsupervised Learning B.2 Dimensionality Reduction. Lars Schmidt-Thieme, Nicolas Schilling

Machine Learning. B. Unsupervised Learning B.2 Dimensionality Reduction. Lars Schmidt-Thieme, Nicolas Schilling Machine Learning B. Unsupervised Learning B.2 Dimensionality Reduction Lars Schmidt-Thieme, Nicolas Schilling Information Systems and Machine Learning Lab (ISMLL) Institute for Computer Science University

More information

VIBRATION-BASED DAMAGE DETECTION UNDER CHANGING ENVIRONMENTAL CONDITIONS

VIBRATION-BASED DAMAGE DETECTION UNDER CHANGING ENVIRONMENTAL CONDITIONS VIBRATION-BASED DAMAGE DETECTION UNDER CHANGING ENVIRONMENTAL CONDITIONS A.M. Yan, G. Kerschen, P. De Boe, J.C Golinval University of Liège, Liège, Belgium am.yan@ulg.ac.be g.kerschen@ulg.ac.bet Abstract

More information

IOMAC' May Guimarães - Portugal RELATIONSHIP BETWEEN DAMAGE AND CHANGE IN DYNAMIC CHARACTERISTICS OF AN EXISTING BRIDGE

IOMAC' May Guimarães - Portugal RELATIONSHIP BETWEEN DAMAGE AND CHANGE IN DYNAMIC CHARACTERISTICS OF AN EXISTING BRIDGE IOMAC'13 5 th International Operational Modal Analysis Conference 2013 May 13-15 Guimarães - Portugal RELATIONSHIP BETWEEN DAMAGE AND CHANGE IN DYNAMIC CHARACTERISTICS OF AN EXISTING BRIDGE Takeshi Miyashita

More information

OBJECT DETECTION AND RECOGNITION IN DIGITAL IMAGES

OBJECT DETECTION AND RECOGNITION IN DIGITAL IMAGES OBJECT DETECTION AND RECOGNITION IN DIGITAL IMAGES THEORY AND PRACTICE Bogustaw Cyganek AGH University of Science and Technology, Poland WILEY A John Wiley &. Sons, Ltd., Publication Contents Preface Acknowledgements

More information

STRUCTURAL DAMAGE DETECTION USING SIGNAL-BASED PATTERN RECOGNITION LONG QIAO

STRUCTURAL DAMAGE DETECTION USING SIGNAL-BASED PATTERN RECOGNITION LONG QIAO STRUCTURAL DAMAGE DETECTION USING SIGNAL-BASED PATTERN RECOGNITION by LONG QIAO B.S., Xian University of Architecture and Technology, China, 998 M.S., Texas Tech University, Lubbock, 23 AN ABSTRACT OF

More information

Machine learning for pervasive systems Classification in high-dimensional spaces

Machine learning for pervasive systems Classification in high-dimensional spaces Machine learning for pervasive systems Classification in high-dimensional spaces Department of Communications and Networking Aalto University, School of Electrical Engineering stephan.sigg@aalto.fi Version

More information

OUTPUT-ONLY STATISTICAL TIME SERIES METHODS FOR STRUCTURAL HEALTH MONITORING: A COMPARATIVE STUDY

OUTPUT-ONLY STATISTICAL TIME SERIES METHODS FOR STRUCTURAL HEALTH MONITORING: A COMPARATIVE STUDY 7th European Workshop on Structural Health Monitoring July 8-11, 2014. La Cité, Nantes, France More Info at Open Access Database www.ndt.net/?id=17198 OUTPUT-ONLY STATISTICAL TIME SERIES METHODS FOR STRUCTURAL

More information

Fragility. manner. in terms of. or an EDP. frame. As

Fragility. manner. in terms of. or an EDP. frame. As Development of Fragility Functions for Seismic Damage Assessment Using Kernel Smoothing Methods H. oh & A.S. Kiremidjian Stanford University, USA SUMMARY: Efficient and reliable assessment of structural

More information

Correlation Preserving Unsupervised Discretization. Outline

Correlation Preserving Unsupervised Discretization. Outline Correlation Preserving Unsupervised Discretization Jee Vang Outline Paper References What is discretization? Motivation Principal Component Analysis (PCA) Association Mining Correlation Preserving Discretization

More information

A Data-driven Approach for Remaining Useful Life Prediction of Critical Components

A Data-driven Approach for Remaining Useful Life Prediction of Critical Components GT S3 : Sûreté, Surveillance, Supervision Meeting GdR Modélisation, Analyse et Conduite des Systèmes Dynamiques (MACS) January 28 th, 2014 A Data-driven Approach for Remaining Useful Life Prediction of

More information

STRUCTURAL DAMAGE DETECTION THROUGH CROSS CORRELATION ANALYSIS OF MOBILE SENSING DATA

STRUCTURAL DAMAGE DETECTION THROUGH CROSS CORRELATION ANALYSIS OF MOBILE SENSING DATA 5 th World Conference on Structural Control and Monitoring 5WCSCM-3 STRUCTURAL DAMAGE DETECTION THROUGH CROSS CORRELATION ANALYSIS OF MOBILE SENSING DATA Dapeng Zhu, Xiaohua Yi, Yang Wang School of Civil

More information

Experimental Modal Analysis of a Flat Plate Subjected To Vibration

Experimental Modal Analysis of a Flat Plate Subjected To Vibration American Journal of Engineering Research (AJER) 2016 American Journal of Engineering Research (AJER) e-issn: 2320-0847 p-issn : 2320-0936 Volume-5, Issue-6, pp-30-37 www.ajer.org Research Paper Open Access

More information

State-Estimation Techniques for a Simple 3DOF Structure

State-Estimation Techniques for a Simple 3DOF Structure State-Estimation Techniques for a Simple 3DOF Structure Alex Mead Zeshi Zheng I. Abstract Structural health monitoring (SHM) is a relatively new field of study in structural engineering, with the goal

More information

OPTIMIZATION OF RESPONSE SIMULATION FOR LOSS ESTIMATION USING PEER S METHODOLOGY

OPTIMIZATION OF RESPONSE SIMULATION FOR LOSS ESTIMATION USING PEER S METHODOLOGY 13 th World Conference on Earthquake Engineering Vancouver, B.C., Canada August 1-6, 2004 Paper No. 1066 OPTIMIZATION OF RESPONSE SIMULATION FOR LOSS ESTIMATION USING PEER S METHODOLOGY Hesameddin ASLANI

More information

Parametric Study of Self-Centering Concentrically- Braced Frames in Response to Earthquakes

Parametric Study of Self-Centering Concentrically- Braced Frames in Response to Earthquakes The University of Akron IdeaExchange@UAkron Honors Research Projects The Dr. Gary B. and Pamela S. Williams Honors College Spring 2015 Parametric Study of Self-Centering Concentrically- Braced Frames in

More information

Damage detection in a reinforced concrete slab using outlier analysis

Damage detection in a reinforced concrete slab using outlier analysis Damage detection in a reinforced concrete slab using outlier analysis More info about this article: http://www.ndt.net/?id=23283 Abstract Bilal A. Qadri 1, Dmitri Tcherniak 2, Martin D. Ulriksen 1 and

More information

WASHINGTON UNIVERSITY SEVER INSTITUTE OF TECHNOLOGY DEPARTMENT OF CIVIL ENGINEERING ABSTRACT

WASHINGTON UNIVERSITY SEVER INSTITUTE OF TECHNOLOGY DEPARTMENT OF CIVIL ENGINEERING ABSTRACT WASHINGTON UNIVERSITY SEVER INSTITUTE OF TECHNOLOGY DEPARTMENT OF CIVIL ENGINEERING ABSTRACT TWO STRUCTURAL HEALTH MONITORING STRATEGIES BASED ON GLOBAL ACCELERATION RESPONSES: DEVELOPMENT, IMPLEMENTATION,

More information

PCA & ICA. CE-717: Machine Learning Sharif University of Technology Spring Soleymani

PCA & ICA. CE-717: Machine Learning Sharif University of Technology Spring Soleymani PCA & ICA CE-717: Machine Learning Sharif University of Technology Spring 2015 Soleymani Dimensionality Reduction: Feature Selection vs. Feature Extraction Feature selection Select a subset of a given

More information

Damage detection of shear connectors in composite bridges under operational conditions

Damage detection of shear connectors in composite bridges under operational conditions Southern Cross University epublications@scu 23rd Australasian Conference on the Mechanics of Structures and Materials 214 Damage detection of shear connectors in composite bridges under operational conditions

More information

FACTOR ANALYSIS AND MULTIDIMENSIONAL SCALING

FACTOR ANALYSIS AND MULTIDIMENSIONAL SCALING FACTOR ANALYSIS AND MULTIDIMENSIONAL SCALING Vishwanath Mantha Department for Electrical and Computer Engineering Mississippi State University, Mississippi State, MS 39762 mantha@isip.msstate.edu ABSTRACT

More information

STRUCTURAL PARAMETERS IDENTIFICATION BASED ON DIFFERENTIAL EVOLUTION AND PARTICLE SWARM OPTIMIZATION

STRUCTURAL PARAMETERS IDENTIFICATION BASED ON DIFFERENTIAL EVOLUTION AND PARTICLE SWARM OPTIMIZATION STRUCTURAL PARAMETERS IDENTIFICATION BASED ON DIFFERENTIAL EVOLUTION AND PARTICLE SWARM OPTIMIZATION Wang Yanwei MEE08167 Supervisor : Xue Songtao Tang Hesheng ABSTRACT Civil structures always suffer many

More information

Heeyoul (Henry) Choi. Dept. of Computer Science Texas A&M University

Heeyoul (Henry) Choi. Dept. of Computer Science Texas A&M University Heeyoul (Henry) Choi Dept. of Computer Science Texas A&M University hchoi@cs.tamu.edu Introduction Speaker Adaptation Eigenvoice Comparison with others MAP, MLLR, EMAP, RMP, CAT, RSW Experiments Future

More information

Sparse orthogonal factor analysis

Sparse orthogonal factor analysis Sparse orthogonal factor analysis Kohei Adachi and Nickolay T. Trendafilov Abstract A sparse orthogonal factor analysis procedure is proposed for estimating the optimal solution with sparse loadings. In

More information

Multiple damage detection in beams in noisy conditions using complex-wavelet modal curvature by laser measurement

Multiple damage detection in beams in noisy conditions using complex-wavelet modal curvature by laser measurement Multiple damage detection in beams in noisy conditions using complex-wavelet modal curvature by laser measurement W. Xu 1, M. S. Cao 2, M. Radzieński 3, W. Ostachowicz 4 1, 2 Department of Engineering

More information

DIMENSION REDUCTION AND CLUSTER ANALYSIS

DIMENSION REDUCTION AND CLUSTER ANALYSIS DIMENSION REDUCTION AND CLUSTER ANALYSIS EECS 833, 6 March 2006 Geoff Bohling Assistant Scientist Kansas Geological Survey geoff@kgs.ku.edu 864-2093 Overheads and resources available at http://people.ku.edu/~gbohling/eecs833

More information

Introduction to Biomedical Engineering

Introduction to Biomedical Engineering Introduction to Biomedical Engineering Biosignal processing Kung-Bin Sung 6/11/2007 1 Outline Chapter 10: Biosignal processing Characteristics of biosignals Frequency domain representation and analysis

More information

Bearing fault diagnosis based on EMD-KPCA and ELM

Bearing fault diagnosis based on EMD-KPCA and ELM Bearing fault diagnosis based on EMD-KPCA and ELM Zihan Chen, Hang Yuan 2 School of Reliability and Systems Engineering, Beihang University, Beijing 9, China Science and Technology on Reliability & Environmental

More information

A Simple Implementation of the Stochastic Discrimination for Pattern Recognition

A Simple Implementation of the Stochastic Discrimination for Pattern Recognition A Simple Implementation of the Stochastic Discrimination for Pattern Recognition Dechang Chen 1 and Xiuzhen Cheng 2 1 University of Wisconsin Green Bay, Green Bay, WI 54311, USA chend@uwgb.edu 2 University

More information

Input-Output Peak Picking Modal Identification & Output only Modal Identification and Damage Detection of Structures using

Input-Output Peak Picking Modal Identification & Output only Modal Identification and Damage Detection of Structures using Input-Output Peak Picking Modal Identification & Output only Modal Identification and Damage Detection of Structures using Time Frequency and Wavelet Techniquesc Satish Nagarajaiah Professor of Civil and

More information

Grandstand Terraces. Experimental and Computational Modal Analysis. John N Karadelis

Grandstand Terraces. Experimental and Computational Modal Analysis. John N Karadelis Grandstand Terraces. Experimental and Computational Modal Analysis. John N Karadelis INTRODUCTION Structural vibrations caused by human activities are not known to be particularly damaging or catastrophic.

More information

Spectral Methods for Subgraph Detection

Spectral Methods for Subgraph Detection Spectral Methods for Subgraph Detection Nadya T. Bliss & Benjamin A. Miller Embedded and High Performance Computing Patrick J. Wolfe Statistics and Information Laboratory Harvard University 12 July 2010

More information

Real Time Face Detection and Recognition using Haar - Based Cascade Classifier and Principal Component Analysis

Real Time Face Detection and Recognition using Haar - Based Cascade Classifier and Principal Component Analysis Real Time Face Detection and Recognition using Haar - Based Cascade Classifier and Principal Component Analysis Sarala A. Dabhade PG student M. Tech (Computer Egg) BVDU s COE Pune Prof. Mrunal S. Bewoor

More information

CS229 Final Project. Wentao Zhang Shaochuan Xu

CS229 Final Project. Wentao Zhang Shaochuan Xu CS229 Final Project Shale Gas Production Decline Prediction Using Machine Learning Algorithms Wentao Zhang wentaoz@stanford.edu Shaochuan Xu scxu@stanford.edu In petroleum industry, oil companies sometimes

More information

1.1 OBJECTIVE AND CONTENTS OF THE BOOK

1.1 OBJECTIVE AND CONTENTS OF THE BOOK 1 Introduction 1.1 OBJECTIVE AND CONTENTS OF THE BOOK Hysteresis is a nonlinear phenomenon exhibited by systems stemming from various science and engineering areas: under a low-frequency periodic excitation,

More information

VIBRATION PROBLEMS IN ENGINEERING

VIBRATION PROBLEMS IN ENGINEERING VIBRATION PROBLEMS IN ENGINEERING FIFTH EDITION W. WEAVER, JR. Professor Emeritus of Structural Engineering The Late S. P. TIMOSHENKO Professor Emeritus of Engineering Mechanics The Late D. H. YOUNG Professor

More information

Anomaly Detection for the CERN Large Hadron Collider injection magnets

Anomaly Detection for the CERN Large Hadron Collider injection magnets Anomaly Detection for the CERN Large Hadron Collider injection magnets Armin Halilovic KU Leuven - Department of Computer Science In cooperation with CERN 2018-07-27 0 Outline 1 Context 2 Data 3 Preprocessing

More information

Linear Regression and Discrimination

Linear Regression and Discrimination Linear Regression and Discrimination Kernel-based Learning Methods Christian Igel Institut für Neuroinformatik Ruhr-Universität Bochum, Germany http://www.neuroinformatik.rub.de July 16, 2009 Christian

More information

Structural Health Monitoring Using Statistical Pattern Recognition Techniques

Structural Health Monitoring Using Statistical Pattern Recognition Techniques Hoon Sohn Engineering Sciences & Applications Division, Engineering Analysis Group, M/S C926 e-mail: sohn@lanl.gov Charles R. Farrar Engineering Sciences & Applications Division, Engineering Analysis Group,

More information

Chemometrics: Classification of spectra

Chemometrics: Classification of spectra Chemometrics: Classification of spectra Vladimir Bochko Jarmo Alander University of Vaasa November 1, 2010 Vladimir Bochko Chemometrics: Classification 1/36 Contents Terminology Introduction Big picture

More information

Principal Component Analysis

Principal Component Analysis Principal Component Analysis Anders Øland David Christiansen 1 Introduction Principal Component Analysis, or PCA, is a commonly used multi-purpose technique in data analysis. It can be used for feature

More information

In order to compare the proteins of the phylogenomic matrix, we needed a similarity

In order to compare the proteins of the phylogenomic matrix, we needed a similarity Similarity Matrix Generation In order to compare the proteins of the phylogenomic matrix, we needed a similarity measure. Hamming distances between phylogenetic profiles require the use of thresholds for

More information

Issues and Techniques in Pattern Classification

Issues and Techniques in Pattern Classification Issues and Techniques in Pattern Classification Carlotta Domeniconi www.ise.gmu.edu/~carlotta Machine Learning Given a collection of data, a machine learner eplains the underlying process that generated

More information

A Novel Vibration-Based Two-Stage Bayesian System Identification Method

A Novel Vibration-Based Two-Stage Bayesian System Identification Method A Novel Vibration-Based Two-Stage Bayesian System Identification Method *Feng-Liang Zhang 1) and Siu-Kui Au 2) 1) Research Institute of Structural Engineering and isaster Reduction, Tongji University,

More information

ANNEX A: ANALYSIS METHODOLOGIES

ANNEX A: ANALYSIS METHODOLOGIES ANNEX A: ANALYSIS METHODOLOGIES A.1 Introduction Before discussing supplemental damping devices, this annex provides a brief review of the seismic analysis methods used in the optimization algorithms considered

More information

Induction Motor Bearing Fault Detection with Non-stationary Signal Analysis

Induction Motor Bearing Fault Detection with Non-stationary Signal Analysis Proceedings of International Conference on Mechatronics Kumamoto Japan, 8-1 May 7 ThA1-C-1 Induction Motor Bearing Fault Detection with Non-stationary Signal Analysis D.-M. Yang Department of Mechanical

More information

Forecast comparison of principal component regression and principal covariate regression

Forecast comparison of principal component regression and principal covariate regression Forecast comparison of principal component regression and principal covariate regression Christiaan Heij, Patrick J.F. Groenen, Dick J. van Dijk Econometric Institute, Erasmus University Rotterdam Econometric

More information

Discriminant analysis and supervised classification

Discriminant analysis and supervised classification Discriminant analysis and supervised classification Angela Montanari 1 Linear discriminant analysis Linear discriminant analysis (LDA) also known as Fisher s linear discriminant analysis or as Canonical

More information

ECE 521. Lecture 11 (not on midterm material) 13 February K-means clustering, Dimensionality reduction

ECE 521. Lecture 11 (not on midterm material) 13 February K-means clustering, Dimensionality reduction ECE 521 Lecture 11 (not on midterm material) 13 February 2017 K-means clustering, Dimensionality reduction With thanks to Ruslan Salakhutdinov for an earlier version of the slides Overview K-means clustering

More information

Effect of temperature on the accuracy of predicting the damage location of high strength cementitious composites with nano-sio 2 using EMI method

Effect of temperature on the accuracy of predicting the damage location of high strength cementitious composites with nano-sio 2 using EMI method Effect of temperature on the accuracy of predicting the damage location of high strength cementitious composites with nano-sio 2 using EMI method J.S Kim 1), S. Na 2) and *H.K Lee 3) 1), 3) Department

More information