Vibration-based damage detection in an aircraft wing scaled model using principal component analysis and pattern recognition

I. Trendafilova, Matthew P. Cartmell, Wieslaw Ostachowicz, Matthew Cartmell

Research output: Contribution to journalArticlepeer-review

77 Citations (Scopus)
187 Downloads (Pure)


This study deals with vibration-based fault detection in structures and suggests a viable methodology based on principal component analysis (PCA) and a simple pattern recognition (PR) method. The frequency response functions (FRFs) of the healthy and the damaged structure are used as initial data. A PR procedure based on the nearest neighbour principle is applied to recognise between the categories of the damaged and the healthy wing data. A modified PCA method is suggested here, which not only reduces the dimensionality of the FRFs but in addition makes the PCA transformed data from the two categories more differentiable. It is applied to selected frequency bands of FRFs which permits the reduction of the PCA transformed FRFs to two new variables, which are used as damage features. In this study, the methodology is developed and demonstrated using the vibration response of a scaled aircraft wing simulated by a finite element (FE) model. The suggested damage detection methodology is based purely on the analysis of the vibration response of the structure. This makes it quite generic and permits its potential development and application for measured vibration data from real aircraft wings as well as for other real and complex structures.
Original languageEnglish
Pages (from-to)560-566
Number of pages6
JournalJournal of Sound and Vibration
Issue number3-5
Publication statusPublished - 17 Jun 2008


  • aircraft wing
  • damage detection
  • vibration-based health monitoring
  • principal component analysis
  • pattern recognition methods


Dive into the research topics of 'Vibration-based damage detection in an aircraft wing scaled model using principal component analysis and pattern recognition'. Together they form a unique fingerprint.

Cite this