Khan, Abdul Malik (1994) Fourier descriptor approach for defect analysis in ultrasonic nondestructive evaluation. Masters thesis, King Fahd University of Petroleum and Minerals.
Increasing emphasis on reliability and cost effective methods of production has motivated a great deal of research in the area of Non-destructive testing (NDT). Among the several methods involving a variety of physical phenomenon, ultrasonic technique has begun to play important role as NDT tools. One such application described in this thesis involves the use of defect describing function in order to non-destructively classify defects. A new technique is proposed to extract features from defect describing function in order to characterize any hidden flaw. Fourier Descriptor technique is used to formulate the input output describinig function. Features of the defect describing function can be considered as the defect signature. Motivated by the need to improve the accuracy of defect classification, the author under took an exhaustive study of the time domain signals. This observation clearly outlined the need to recognize and determine the sub-classes of features. This thesis describes the use of Fourier descriptors for characterizing closed input-output describing functions. In order to improve the classification error rate and facilitate identification of a larger class of defects, eight different systems with different signal to noise ratios were considered. Experimental verification was done by extracting unique set of features from experimental data of different systems. The signals were classified correctly in all cases.
|Item Type:||Thesis (Masters)|
|Divisions:||College Of Engineering Sciences > Electrical Engineering Dept|
|Creators:||Khan, Abdul Malik|
|Committee Advisor:||Yamani, Ahmed|
|Committee Members:||Halawani, T. H. and Biyari, Khaled H. and Bettayeb, Mammar and Al-Akhdhar, Syed Zaki|
|Deposited By:||KFUPM ePrints Admin|
|Deposited On:||22 Jun 2008 16:55|
|Last Modified:||25 Apr 2011 10:00|
Repository Staff Only: item control page