On the Generalizability of Time-of-Flight Convolutional Neural Networks for Noninvasive Acoustic Measurements
Abhishek Saini1, John James Greenhall1, Eric Sean Davis1
1Los Alamos National Laboratory, Los Alamos, NM 87544, USA.
Sensors (Basel, Switzerland)
|June 19, 2024
Summary
Convolutional neural networks (CNNs) can accurately measure acoustic time-of-flight (ToF) in complex NDE scenarios. This study confirms CNN generalizability with limited data, showing robust performance across various container sizes and noisy conditions.
Area of Science:
- Non-destructive evaluation (NDE)
- Acoustic wave propagation
- Machine learning applications
Background:
- Bulk wave acoustic time-of-flight (ToF) measurements are often hindered by guided waves in container walls, especially at low frequencies.
- Convolutional neural networks (CNNs) show promise for accurate ToF determination in challenging non-destructive evaluation (NDE) conditions.
- The generalizability of CNNs for ToF prediction, particularly with limited training data, remains an underexplored area.
Purpose of the Study:
- To analyze the generalizability of CNNs for acoustic time-of-flight (ToF) prediction in pipes and closed containers.
- To investigate the impact of training dataset size and data parameter distribution on CNN performance.
- To assess the robustness and broader applicability of ToF-CNN models.
Main Methods:
- Investigated CNN performance variations based on training dataset size and parameters (container dimensions, material properties).
- Trained models on diverse datasets (small and large containers) to understand generalizability requirements.
- Augmented data with noise to simulate real-world experimental conditions.
Main Results:
- Sufficient and representative training data is crucial for spanning the input space and achieving accurate ToF prediction.
- CNNs trained on smaller container datasets demonstrated more stable results compared to those trained on larger containers.
- The trained model exhibited excellent accuracy in predicting ToF for different sound speed mediums and performed robustly with noisy data.
Conclusions:
- CNNs offer a generalizable approach for accurate acoustic time-of-flight (ToF) prediction in non-destructive evaluation (NDE).
- Model generalizability is enhanced by ensuring training data adequately represents the full range of input parameters.
- The proposed CNN methodology shows potential for expanded applications in acoustic ToF analysis across various scenarios.
Related Concept Videos
01:26Linear Approximation in Frequency Domain
Linear systems are characterized by two main properties: superposition and homogeneity. Superposition allows the response to multiple inputs to be the sum of the responses to each individual input. Homogeneity ensures that scaling an input by a scalar results in the response being scaled by the same scalar.
In contrast, nonlinear systems do not inherently possess these properties. However, for small deviations around an operating point, a nonlinear system can often be approximated as linear.
In contrast, nonlinear systems do not inherently possess these properties. However, for small deviations around an operating point, a nonlinear system can often be approximated as linear.
01:21Linear Approximation in Time Domain
Nonlinear systems often require sophisticated approaches for accurate modeling and analysis, with state-space representation being particularly effective. This method is especially useful for systems where variables and parameters vary with time or operating conditions, such as in a simple pendulum or a translational mechanical system with nonlinear springs.
For a simple pendulum with a mass evenly distributed along its length and the center of mass located at half the pendulum's length, the...
For a simple pendulum with a mass evenly distributed along its length and the center of mass located at half the pendulum's length, the...