Maximum a posteriori deconvolution of sparse ultrasonic signals using genetic optimization.

Area of Science:

• Materials Science
• Signal Processing
• Non-Destructive Testing

Background:

• Ultrasonic non-destructive testing (NDT) faces challenges with closely spaced reflections masking defects in layered structures.
• Signal deconvolution is crucial for interpreting NDT data, similar to seismic exploration challenges.
• Detecting disbonds in layered aluminum structures is a key application requiring advanced signal processing.

Purpose of the Study:

• To develop an effective deconvolution method for ultrasonic NDT signals.
• To improve the detection and interpretation of closely spaced reflections in layered materials.
• To apply a novel genetic algorithm for sparse spike sequence estimation.

Main Methods:

• Modeling signal generation using the Bernoulli-Gaussian (BG) distribution.
• Employing Maximum A Posteriori (MAP) estimation for reflection sequence recovery.
• Utilizing a modified genetic algorithm to optimize the non-convex MAP criterion.

Main Results:

• The proposed genetic algorithm successfully optimized the MAP criterion for BG deconvolution.
• Tested on simulated and real ultrasonic data, the algorithm demonstrated improved signal interpretation.
• Evaluation using Probability of Detection (PD) and Probability of False Alarm (PFA) showed significant enhancements.

Conclusions:

• The genetic algorithm-based deconvolution offers a robust solution for NDT signal analysis.
• This approach enhances the capability to detect disbonds and interpret complex reflection patterns.
• The modified genetic algorithm improves the efficiency and reliability of deconvolution in ultrasonic NDT.