Defects Detection and Classification in Low Conductive Materials using Artificial Neural Network and Principal Component Analysis

  • Yusra A. Abdulrahman

Student thesis: Doctoral Thesis


Infrared Thermography (IRT thereafter), in comparison to other nondestructive testing approaches, is increasing its penetration into new industrial applications, mainly due to its contact-less nature, area coverage, and fast inspection modes. IRT thermograms can result from passive and/or active stimulation techniques and need to be pre and post-processed to yield qualitative information about embedded defectives in the host structure. In this research, several new techniques are investigated and supplemented to yield quantitative information that relates to the defectives' depth. Moreover, the presented dissertation will offer insights on the autonomous detection with little to no human intervention or major calibration effort. The research will address the use of Principal Component Analysis or PCA as a means to digest and consolidate the large sets of imagery data from the physical space; and at the same time, produce enhanced thermograms. The research will also discuss the limitation of the PCA as a mathematical tool that does not keep the peak contrast and its associated timestamp, as it processes the IRT images; thus losing the physics of the heat transfer inside the host material. This study will also aim at researching and using a reliable classification/learning technique for the defectives found in a low conductive host, which is Carbon Fiber Reinforced Plastics Composite CFRP, via an Artificial Neural Network ANN novel approach. The presented ANN scheme is compared to the PCA outcome and is implemented on artificially designed samples. Supplementary studies from this research will also show the use of a feedforward ANN, to present full autonomous detection and quantification.
Date of AwardMay 2020
Original languageAmerican English
SupervisorMohammad Omar (Supervisor)


  • Non-Destructive Testing
  • Infrared Thermography
  • 3D printing
  • Principal Component Analysis
  • Artificial Neural Network.

Cite this