On the Generalizability of Time-of-Flight Convolutional Neural Networks for Noninvasive Acoustic Measurements.

Bulk wave acoustic time-of-flight (ToF) measurements in pipes and closed containers can be hindered by guided waves with similar arrival times propagating in the container wall, especially when a low excitation frequency is used to mitigate sound attenuation from the material. Convolutional neural networks (CNNs) have emerged as a new paradigm for obtaining accurate ToF in non-destructive evaluation (NDE) and have been demonstrated for such complicated conditions. However, the generalizability of ToF-CNNs has not been investigated. In this work, we analyze the generalizability of the ToF-CNN for broader applications, given limited training data. We first investigate the CNN performance with respect to training dataset size and different training data and test data parameters (container dimensions and material properties). Furthermore, we perform a series of tests to understand the distribution of data parameters that need to be incorporated in training for enhanced model generalizability. This is investigated by training the model on a set of small- and large-container datasets regardless of the test data. We observe that the quantity of data partitioned for training must be of a good representation of the entire sets and sufficient to span through the input space. The result of the network also shows that the learning model with the training data on small containers delivers a sufficiently stable result on different feature interactions compared to the learning model with the training data on large containers. To check the robustness of the model, we tested the trained model to predict the ToF of different sound speed mediums, which shows excellent accuracy. Furthermore, to mimic real experimental scenarios, data are augmented by adding noise. We envision that the proposed approach will extend the applications of CNNs for ToF prediction in a broader range.

Genetic Algorithm-Wavelet Transform Feature Extraction for Data-Driven Acoustic Resonance Spectroscopy.

Acoustic resonance spectroscopy (ARS) enables highly accurate measurement of the properties (geometry/material) of a structure based on the structure's natural vibrational resonances. In general, measuring a specific property in multibody structures presents a significant challenge due to the complex overlapping peaks within the resonance spectrum. We present a technique for extracting useful features from a complex spectrum by isolating resonance peaks that are sensitive to the measured property and insensitive to other properties (noise peaks). We isolate specific peaks by selecting frequency regions of interest and performing wavelet transformation, where the frequency regions and wavelet scales are tuned via a genetic algorithm. This contrasts greatly from the traditional wavelet transformation/decomposition techniques, which use a large number of wavelets at different scales to represent the signal, including the noise peaks, and results in a large feature size, thus decreasing machine learning (ML) generalizability. We provide a detailed description of the technique and demonstrate the feature extraction technique, for example, regression and classification problems. We observe reductions of 95% and 40% in regression and classification errors, respectively, when using the genetic algorithm/wavelet transform feature extraction, compared to using no feature extraction, or using wavelet decomposition, which is common in optical spectroscopy. The feature extraction has potential to significantly increase the accuracy of spectroscopy measurements based on a wide range of ML techniques. This would have significant implications for ARS, as well as other data-driven methods for other types of spectroscopy, e.g., optical.

Noninvasive Acoustic Measurements in Cylindrical Shell Containers.

Acoustic time-of-flight (ToF) measurements enable noninvasive material characterization, acoustic imaging, and defect detection and are commonly used in industrial process control, biomedical devices, and national security. When characterizing a fluid contained in a cylinder or pipe, ToF measurements are hampered by guided waves, which propagate around the cylindrical shell walls and obscure the waves propagating through the interrogated fluid. We present a technique for overcoming this limitation based on a broadband linear chirp excitation and cross correlation detection. By using broadband excitation, we exploit the dispersion of the guided waves, wherein different frequencies propagate at different velocities, thus distorting the guided wave signal while leaving the bulk wave signal in the fluid unperturbed. We demonstrate the measurement technique experimentally and using numerical simulation. We characterize the technique performance in terms of measurement error, signal-to-noise-ratio, and resolution as a function of the linear chirp center frequency and bandwidth. We discuss the physical phenomena behind the guided bulk wave interactions and how to utilize these phenomena to optimize the measurements in the fluid.