Loading...

Robust Similarity Measure for Acoustical Leak Detection in the Presence of Different Noise

Ahmadi, Ali Mohammad | 2017

986 Viewed
  1. Type of Document: Ph.D. Dissertation
  2. Language: Farsi
  3. Document No: 50424 (04)
  4. University: Sharif University of Technology
  5. Department: Physics
  6. Advisor(s): Amjadi, Ahmad; Bahrampour, Alireza
  7. Abstract:
  8. Precise leak detection by acoustic signal analysis calls for robust similarity measures to estimate the time delay between recorded leakage signals and minimize the probability of false alarms, in the face of dispersive propagation, multiple reflections, and unknown correlated, uncorrelated and impulsive background noise. Providing evidence that higher order modes and multi-reflected signals behave like sets of correlated noise, this thesis uses a regression model to optimize residual complexity in the presence of both correlated and uncorrelated noise. This optimized residual complexity (ORC) is highly robust since it takes into account both the level and complexity of noise and signal, and it optimizes the trade-off between sensitivity to the energy or to the complexity of residual signal, dependenting on the noise mixture.
    The lower complexity of the dispersive modes and multiple reflections, compared to the complexity of the plane mode, points to the robustness of ORC against multiple reflections and dispersion. Experimental investigations using recorded sounds of gas leaking from a pipe confirm the robustness of ORC against multiple reflections. Numerical simulations also show robustness against dispersive modes, even when they disturb the linearity of the cross-spectrum phase. Comparisons with other maximum-likelihood methods—mutual information, cross correlation, and residual complexity—underline the general advantages of ORC in terms of robustness in the presence of reflection and dispersion, against both correlated and uncorrelated noise, and to short signals.
    To become robust against high-energy correlated and impulsive noise too, the other proposed similarity measure—wavelet based optimized residual complexity (WORC)— employs wavelet basis in order to reduce the complexity of mixture of correlated and impulsive noise. In order to compare the resolution and percentage of false alarms of WORC with those of optimized residual complexity and cross correlation, we used both experimental and simulated leakage signals of a gas pipe. The results demonstrate the superiority of WORC in term of robustness in the presence of mixture of correlated, uncorrelated, and impulsive noise
  9. Keywords:
  10. Dispersion ; Maximum Likelihood Estimation ; Correlation ; Uncorrelated Independant Noise Sources ; Impulsive Noise Removal ; Nonlinear Regression ; Acoustical Leak Detection ; Time Delay Estimation ; Optimized Residual Complexity

 Digital Object List

 Bookmark

No TOC