Feature selection techniques for identifying the most relevant damage indices in SHM using Guided Waves

Feature selection techniques aim to evaluate feature’s importance and select the most relevant ones. This paper concerns the selection of features in order to perform a reliable Structural Health Monitoring by means of ultrasonic guided waves technique. The current case of study deals with the health monitoring of pipelines. A corrosion-like defect was machined in a full-scale tube and then its size was increased in five steps. Their cross-section areas (CSA) go from less than 1% to around 4.5%. To get a high accur acy, a 3D laser scanner was used to measure these CSAs. Many signal features were extracted from the ultrasonic signals. An algorithm, called sequential forward feature selection, was applied on these features to select the most discriminating ones. For the ease of the reader, a background of feature selection algorithms is presented. Damage detection procedure, basing on the Mahalanobis distance, is described. The obtained results show that all defect steps were successfully detected even the smallest one.

[1]  Keith Worden,et al.  An introduction to structural health monitoring , 2007, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[2]  Ron Kohavi,et al.  Wrappers for Feature Subset Selection , 1997, Artif. Intell..

[3]  Wilfried N. Gansterer,et al.  On the Relationship Between Feature Selection and Classification Accuracy , 2008, FSDM.

[4]  S. Mallat.  A wavelet tour of signal processing , 1998 .

[5]  Gavin C. Cawley,et al.  On Over-fitting in Model Selection and Subsequent Selection Bias in Performance Evaluation , 2010, J. Mach. Learn. Res..

[6]  J. Rose,et al.  Flaw Classification Potential in Tubing with Guided Waves , 1995 .

[7]  K. Worden,et al.  The application of machine learning to structural health monitoring , 2007, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[8]  Huan Liu,et al.  Feature selection for classification: A review , 2014 .

[9]  Hoon Sohn,et al.  Effects of environmental and operational variability on structural health monitoring , 2007, Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.

[10]  Huan Liu,et al.  Feature Selection for Classification , 1997, Intell. Data Anal..

[11]  Michel Verleysen,et al.  The Curse of Dimensionality in Data Mining and Time Series Prediction , 2005, IWANN.

[12]  Syed Ismail Shah,et al.  Techniques to Obtain Good Resolution and Concentrated Time-Frequency Distributions: A Review , 2009, EURASIP J. Adv. Signal Process..

[13]  Amparo Alonso-Betanzos,et al.  Filter Methods for Feature Selection - A Comparative Study , 2007, IDEAL.

[14]  Xiaohui Zhang,et al.  Flaw classification in ultrasonic guided waves signal using Wavelet Transform and PNN classifier , 2011, 2011 International Conference on Wireless Communications and Signal Processing (WCSP).

[15]  Hendrik Blockeel,et al.  On estimating model accuracy with repeated cross-validation , 2012 .

[16]  Andrew R. Webb,et al.  Statistical Pattern Recognition , 1999 .

[17]  Abdollah Bagheri,et al.  Outlier Analysis and Artificial Neural Network for the Noncontact Nondestructive Evaluation of Immersed Plates , 2015 .

[18]  James H. Garrett,et al.  A data-driven framework for ultrasonic structural health monitoring of pipes , 2012 .

[19]  Francesco Lanza di Scalea,et al.  Discrete wavelet transform to improve guided-wave-based health monitoring of tendons and cables , 2004, SPIE Smart Structures and Materials + Nondestructive Evaluation and Health Monitoring.