Robust methods for outlier detection and regression for SHM applications

Dervilis, Nikolaos; Antoniadou, Ifigeneia; Barthorpe, R. J.; Cross, Elizabeth J.; Worden, Keith

doi:10.1504/ijsmss.2015.078354

Cited by 13 publications

(24 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Principal component analysis is a data analysis technique that re-expresses the original data in a new basis where the information is arranged in terms of maximal variance and minimal redundancy [64,65]. In the field of SHM and novelty detection, it is important to characterize those changes occurring under normal operation, as they may compromise the efficiency of the assessment method [43,51]. Further details of this procedure are available in [66][67][68], where authors present various applications for SHM.…”

Section: Principal Component Analysismentioning

confidence: 99%

“…Some recently applied algorithms include artificial neural networks (ANNs) [36,37], support vector machines [38,39], or k-nearest neighbor [40]. There are also several works that employ unsupervised machine learning approaches to civil engineering applications [41][42][43][44][45], but these are more powerful in supervised learning contexts, where there is full knowledge about the outcome of each training sample [11].…”

Section: Introductionmentioning

confidence: 99%

“…When working with civil engineering structures under service, we often employ unsupervised learning techniques due to the lack of damaged data [11,43]. This situation leads to the implementation of novelty detection algorithms, which detect deviations from what it is considered the reference behavior but are unable to characterize the type and extent of the damage [46].…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Bearing assessment tool for longitudinal bridge performance

Fernández-Navamuel

Zamora-Sánchez

Alvear

et al. 2020

J Civil Struct Health Monit

View full text Add to dashboard Cite

This work provides an unsupervised learning approach based on a single-valued performance indicator to monitor the global behavior of critical components in a viaduct, such as bearings. We propose an outlier detection method for longitudinal displacements to assess the behavior of a singular asymmetric prestressed concrete structure with a 120 m high central pier acting as a fixed point. We first show that the available long-term horizontal displacement measurements recorded during the undamaged state exhibit strong correlations at the different locations of the bearings. Thus, we combine measurements from four sensors to design a robust performance indicator that is only weakly affected by temperature variations after the application of principal component analysis. We validate the method and show its efficiency against false positives and negatives using several metrics: accuracy, precision, recall, and F1 score. Due to its unsupervised learning scope, the proposed technique is intended to serve as a real-time supervision tool that complements maintenance inspections. It aims to provide support for the prioritization and postponement of maintenance actions in bridge management.

show abstract

Section: Principal Component Analysismentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Bearing assessment tool for longitudinal bridge performance

Fernández-Navamuel

Zamora-Sánchez

Alvear

et al. 2020

J Civil Struct Health Monit

View full text Add to dashboard Cite

show abstract

“…In [3], Mahalanobis distances were calculated between the baseline condition of a structure and the condition that is tested for the existence of damage. More complicated and more robust approaches for detecting the existence of damage are given in [4]. Another novelty measure that indicates existence of damage is given in [5]; it is defined using autoencoders in order to describe the manifold to which the data of the normal condition of the structure belong.…”

Section: What Kind Of Damage Is Present (Type/classification)?mentioning

confidence: 99%

A Neat Approach to Structural Health Monitoring

Tsialiamanis¹,

Wagg²,

Dervilis³

et al. 2020

XI International Conference on Structural Dynamics

Self Cite

View full text Add to dashboard Cite

In the current paper, an application of the neuroevolution of augmenting topologies (NEAT) algorithm is considered in a structural health monitoring (SHM) application. The algorithm is a variation of genetic algorithms, applied in neural networks, and has the goal of optimising both the topology and the weights and biases of a neural network model. The algorithm is applied here to an SHM problem instead of using feedforward neural networks. The algorithm is called to search for the best-fitting topology in the task, which would otherwise be sought through experimenting with the size and number of the layers of the neural network. Having used the algorithm, the accuracy is found to be close to the one achieved using classically trained neural networks. Another aspect of the application is that subnetworks were defined for every damage case of the problem, whose topologies are much simpler than a fullyconnected feedforward neural network. These subnetworks define classification submodels that may be used in different combinations, building models for a subset of damage cases and input features.

show abstract

“…manyexistence modelsare robust towardsoutliers. Dervilis et al (2016) proposed a robust regression for outlier detection by exploring and visualizing structural health monitoring (SHM) data as a tool in investigating and monitoring the characteristics of outlier whilst removing it from the data. Dervilis et al (2014bDervilis et al ( , 2015 and Rousseeuw, Hubert and Aelst (2008) proposed a high-breakdown method that is robust against the outliers where it can deal with a substantial fraction of outliers in the data (Rousseeuw&Hubert, 2013).…”

Section: Notmentioning

confidence: 99%

Comparing the Performance of Winsorize Tree to Other Data Mining Techniques for Cases Involving Outliers

2019

ijrte

View full text Add to dashboard Cite

Winsorize tree is a modified tree that reformed from classification and regression tree (CART). It lays on the strategy of handling and accommodating the outliers simultaneously in all nodes while generating the subsequence branches of tree. Normally, due to the existence of outlier, the accuracy rate of most of the classifiers will be affected. Therefore, we propose winsorize tree which could resist to anomaly data. It protects the originality of the data while performing the splitting process. In this study, winsorize tree was compared to other classifiers. The results obtained from five real datasets indicate that the proposed winsorize tree performs as good as or even better compare to the other data mining techniques based on the misclassification rate.

show abstract

Robust methods for outlier detection and regression for SHM applications

Cited by 13 publications

References 33 publications

Bearing assessment tool for longitudinal bridge performance

Bearing assessment tool for longitudinal bridge performance

A Neat Approach to Structural Health Monitoring

Comparing the Performance of Winsorize Tree to Other Data Mining Techniques for Cases Involving Outliers

Contact Info

Product

Resources

About