Optimal selection of autoregressive model coefficients for early damage detectability with an application to wind turbine blades

Hoell, Simon; Omenzetter, Piotr

doi:10.1016/j.ymssp.2015.09.007

Cited by 39 publications

(15 citation statements)

References 67 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The present study builds on and extends the authors' previous work presented in [8] and [28], where optimal selections of the initial and PCA-transformed time series-based DSFs were studied, respectively. The previous research demonstrated, via tracktable conceptual examples, and validated, through extensive numerical and experimental studies, that, depending on the correlation between the initial DSF components and their sensitivity to damage, different selections and projections of DSFs can be optimal for SDD.…”

Section: Introductionmentioning

confidence: 73%

“…Damage sensitive features (DSFs) are extracted from vibration signals to reduce data volumes and to improve damage detectability. These DSFs are usually multivariate and include modal parameters [6], parametric time series model coefficients [7][8][9][10], and non-parametric time series representations in the frequency domain [11,12], time-frequency domain [13] or time domain. For the latter, amplitudes of cross-correlations [14] and their coefficients [15] or cross-correlations [16] and autocorrelations [17] at the zero lag were all proposed for SDD.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Sequential projection pursuit for optimised vibration-based damage detection in an experimental wind turbine blade

Hoell

Omenzetter

2018

Smart Mater. Struct.

Self Cite

View full text Add to dashboard Cite

To advance the concept of smart structures in large systems, such as wind turbines, it is desirable to be able to detect structural damage early while using minimal instrumentation. Data-driven vibration-based damage detection methods can be competitive in that respect because global vibrational responses encompass the entire structure. Multivariate damage sensitive features (DSFs) extracted from acceleration responses enable to detect changes in a structure via statistical methods. However, even though such DSFs contain information about the structural state, they may not be optimized for the damage detection task. This paper addresses the shortcoming by exploring a DSF projection technique specialized for statistical structural damage detection. High dimensional initial DSFs are projected onto a low-dimensional space for improved damage detection performance and simultaneous computational burden reduction. The technique is based on sequential projection pursuit where the projection vectors are optimized one by one using an advanced evolutionary strategy. The approach is applied to laboratory experiments with a small-scale wind turbine blade under wind-like excitations. Autocorrelation function coefficients calculated from acceleration signals are employed as DSFs. The optimal numbers of projection vectors are identified with the help of a fast forward selection procedure. To benchmark the proposed method, selections of original DSFs as well as principal component analysis scores from these features are additionally investigated. The optimized DSFs are tested for damage detection on previously unseen data from the healthy state and a wide range of damage scenarios. It is demonstrated that using selected subsets of the initial and transformed DSFs improves damage detectability compared to the full set of features. Furthermore, superior results can be achieved by projecting autocorrelation coefficients onto just a single optimized projection vector.

show abstract

Section: Introductionmentioning

confidence: 73%

Section: Introductionmentioning

confidence: 99%

Sequential projection pursuit for optimised vibration-based damage detection in an experimental wind turbine blade

Hoell

Omenzetter

2018

Smart Mater. Struct.

Self Cite

View full text Add to dashboard Cite

show abstract

“…input–output vs output-only), a broad range of representations can be adopted to extract significant DSFs. Considering linear and stationary vibration data, widely used time-invariant linear models are the autoregressive (AR), 18–20 autoregressive with exogenous input (ARX), 21 autoregressive moving average (ARMA), 22,23 autoregressive moving average with exogenous input (ARMAX), 24,25 and autoregressive and autoregressive with exogenous input (ARARX) 26 ones. However, vector-dependent functionally pooled (VFP) models, 27 Gaussian process (GP) time series models, 28 time-varying autoregressive moving average (TV-ARMA), 29 and trigonometric Box–Cox ARMA trend seasonal (TBATS) 30 are suitable for conditions that time series data are exposed to uncertainties, operational and environmental variability, ambient vibration, and seasonal variations.…”

Section: Introductionmentioning

confidence: 99%

Fast unsupervised learning methods for structural health monitoring with large vibration data from dense sensor networks

Entezami

Shariatmadar

Mariani

2019

Structural Health Monitoring

View full text Add to dashboard Cite

Data-driven damage localization is an important step of vibration-based structural health monitoring. Statistical pattern recognition based on the prominent steps of feature extraction and statistical decision-making provides an effective and efficient framework for structural health monitoring. However, these steps may become time-consuming or complex when there are large volumes of vibration measurements acquired by dense sensor networks. To deal with this issue, this study proposes fast unsupervised learning methods for feature extraction through autoregressive modeling and damage localization through a new distance measure called Kullback–Leibler divergence with empirical probability measure. The feature extraction approach consists of an iterative algorithm for order selection and parameter estimation aiming to extract residuals in the training phase and another iterative process aiming to extract residuals only in the monitoring phase. The key feature of the proposed approach is the use of correlated residual samples of the autoregressive model as a new time series at each iteration, rather than handling the measured vibration response of the structure. This is shown to highly reduce the computational burden of order selection and feature extraction; moreover, it effectively provides low-order autoregressive models with uncorrelated residuals. The Kullback–Leibler divergence with empirical probability measure method exploits a segmentation technique to subdivide random data into independent sets and provides a distance metric based on the theory of empirical probability measure with no need to explicitly compute the actual probability distributions at the training and monitoring stages. Numerical and experimental benchmarks are then used to assess accuracy and performance of the proposed methods and compare them with some state-of-the-art approaches. Results show that the proposed approaches are successful in feature extraction and damage localization, with a reduced computational burden.

show abstract

“…On the second step, the fault isolation, is handled via Fuzzy/Bayesian network scheme classifying the kind of fault. Optimal selection of auto regressive model coefficients for early damage detects ability with an application to wind turbine blades was carried out by Hoell and Omenzetter [12].…”

Section: Introductionmentioning

confidence: 99%

Research on Hyper pipes and Voting Feature Intervals Classifier for Condition Monitoring of Wind Turbine Blades using Vibration Signals

2019

IJRTE

View full text Add to dashboard Cite

Renewable energy is viewed as a vital energy field due to the present energy devastations. Among the vital substitutions being considered, wind energy is a strong challenger as a result of its reliability. “To yield wind energy more effectively, the structure of wind turbines has designed bigger, making protection and restoration works difficult. Because of different natural conditions, wind turbine blades are exposed to vibration and it prompts failure. If the failure is not analyzed initially, then it will haste dreadful destruction of the turbine structure. To increase safety perceptions, to decrease down time and to cut down the repeat of unpredictable breakdowns, the wind turbine blades must be examined from time to time to guarantee that they are in great condition. In this paper, a three bladed wind turbine was preferred and using vibration source through statistical features, the state of a wind turbine blade is inspected. The fault classification is carried out using machine learning techniques like hyperpipes (HP) and voting feature intervals (VFI) algorithm. The performance of these algorithms is compared and better algorithm is suggested for fault prediction on wind turbine blades.”

show abstract

Optimal selection of autoregressive model coefficients for early damage detectability with an application to wind turbine blades

Cited by 39 publications

References 67 publications

Sequential projection pursuit for optimised vibration-based damage detection in an experimental wind turbine blade

Sequential projection pursuit for optimised vibration-based damage detection in an experimental wind turbine blade

Fast unsupervised learning methods for structural health monitoring with large vibration data from dense sensor networks

Research on Hyper pipes and Voting Feature Intervals Classifier for Condition Monitoring of Wind Turbine Blades using Vibration Signals

Contact Info

Product

Resources

About