Vibration-based health monitoring approach for composite structures using multivatiate statistical analysis

David Garcia, Irina Trendafilova, Hussein Razzaq Sabah Al-Bugharbee

Research output: Contribution to conferencePaper

74 Downloads (Pure)


In this paper a novel procedure for damage assessment is suggested which is based on singular spectrum analysis (SSA). The main feature of the method is that it applies Principal Component Analysis (PCA) to the lagged time series, obtained from the measured structural vibration response. In this study the methodology is developed for the case of a free decay response. The measured acceleration vectors are transformed into the frequency domain and then used to define a trajectory matrix. The covariance matrix of the trajectory matrix is decomposed into new variables, the Principal Components (PCs). They define a new space of linearly correlated variables onto which the dynamics/motion of the system can be projected. This decomposition is used to uncover oscillation patterns among other purposes. The method is applied and demonstrated for the case of a simple 2-DoF system. To demonstrate its capabilities for damage diagnosis different levels of stiffness reduction are introduced. The first two PCs are used to visually demonstrate the abilities of the methodology. The Mahalanobis distance is used to develop a classification system to detect and localize delamination in the 2-DoF system. The results clearly demonstrate the capabilities of the system to clearly detect and localize damage.
Original languageEnglish
Number of pages8
Publication statusPublished - 8 Jul 2014
Event14th European Workshop on Structura Health Monitoring - Nantes, France
Duration: 8 Jul 201411 Jul 2014


Conference14th European Workshop on Structura Health Monitoring


  • composite materials
  • delamination
  • singular spectrum analysis.
  • statistical pattern recognition
  • damage assessment


Dive into the research topics of 'Vibration-based health monitoring approach for composite structures using multivatiate statistical analysis'. Together they form a unique fingerprint.

Cite this