Measuring predictability in ultrasonic signals: an application to scattering material characterization

Area of Science:

• Materials Science
• Non-Destructive Testing
• Signal Processing

Background:

• Ultrasonic non-destructive testing (NDT) often faces challenges with scattering materials and low signal-to-noise ratios (SNR).
• Traditional methods may require assumptions about material properties like attenuation profiles, limiting their applicability.
• Characterizing scattering media is crucial for accurate material assessment in various industrial applications.

Purpose of the Study:

• To present a novel, non-parametric approach for scattering material characterization using time series predictability.
• To demonstrate the efficacy of this technique in classifying materials based on their scattering properties.
• To validate the method's ability to accurately determine material porosity percentages.

Main Methods:

• Utilized a non-parametric technique from chaos theory to measure the degree of predictability in ultrasonic time series data.
• Applied the technique to 32 cement probes with varying porosity levels.
• Analyzed the relationship between signal predictability and the deterministic component versus noise in the time series.

Main Results:

• The degree of predictability successfully classified cement samples based on their porosity.
• The technique proved effective even in highly scattering media with low SNR.
• Measured predictability accurately indicated the percentage of porosity in the tested cement samples.

Conclusions:

• Time series predictability offers a robust method for scattering material characterization in ultrasonic NDT.
• This approach provides valuable insights into coherent reflections versus incoherent noise within materials.
• The technique demonstrates high accuracy for classifying material porosity, showing significant potential for industrial applications.