Higher-order statistics (HOS)-based deconvolution for ultrasonic nondestructive evaluation (NDE) of materials

(1997) Higher-order statistics (HOS)-based deconvolution for ultrasonic nondestructive evaluation (NDE) of materials. Masters thesis, King Fahd University of Petroleum and Minerals.

PDF (MSc Thesis)

Download (4MB) | Preview


High resolution signal processing techniques involving higher-order statistics (HOS) and artificial neural networks (ANN) which re useful in ultrasonic nondestructive evaluation (NDE) of materials systems subject to additive white Gaussian noise (AWGN) and masking effects of measurement systems used and propagation paths, are investigated in this Thesis. The proposed techniques are: i) a batch-type deconvolution method using the complex bicepstrum algorithm, and ii) automatic ultrasonic defect classification system using a modular learning strategy. Performance evaluation of the proposed methods and comparisons with existing methods are made by means of Monte-Carlo simulations, experimental data and analysis. The first scheme makes use of the complex cepstrum of the third-order cumulants (complex bicepstrum) of the ultrasonic signals. The second scheme based on a modular learning stretegy consisting of three functional blocks, takes into account the nonstationary character of the ultrasonic NDE system and makes use of the " information preserving rule" which allows accurate and reliable classification procedure. It is demonstrated that the proposed techniques perform very efficiently in both white or colored Gaussian and symmetrically-distributed noise-classes at moderate and low signal-to-noise ratios (SNR). Comparisons with existing methods demonstrate improved performance characterized by high resolution properties, robustness and low sensitivity to additive Gaussian noise. However, the improved performance is achieved at the expense of higher computational complexity and data requirements.

Item Type: Thesis (Masters)
Subjects: Engineering
Department: College Of Computer Sciences and Engineering > Information and Computer Science Dept
Depositing User: LAHOUARI GHOUTI
Date Deposited: 18 Jun 2008 08:22
Last Modified: 01 Nov 2019 13:46
URI: https://eprints.kfupm.edu.sa/id/eprint/9143