Probabilistic machine learning aided acoustic emission source localization in mode-I fracture test of multi-layer engineered wood

Prediction of acoustic emission source localization in RC frames using GWO-DNN

This paper presents a high-sensitivity two-dimensional fiber-optic hydrophone designed for the detection and localization of underwater acoustic sources. The device comprises two sensing heads, each incorporating a fiber Bragg grating (FBG) embedded within a customized 3D-printed encapsulation. To enhance acoustic sensitivity, the design utilizes a silicone thin-film coupled with a pyramidal channel that spatially concentrates acoustic energy from the base to the apex, where the FBG is positioned. Incident acoustic pressure induces vibrations in the film, which are amplified by the channel structure, imparting strain on the FBG and resulting in a shift in the Bragg wavelength. The acoustic frequency response is demodulated by converting the overlapping optical power between the sensing and reference gratings into an electrical signal via a photodetector. By arranging the two sensing heads orthogonally, the system effectively determines the direction and angle of the acoustic source. Experimental results show a peak sensitivity of -210.59 dB re 1 V/μPa, with a FWHM of 57.92-66.27 Hz and a figure of merit (FOM) up to 3.64 dB/Hz. In addition, the acoustic-field SNR is approximately 26 dB in the dominant band, and the LOD is 64.19 dB re 1 μPa (10-400 Hz). Experimental validation confirms the hydrophone's high sensitivity and localization accuracy, demonstrating its significant potential for underwater acoustic sensing applications.

Abstract Identifying the spatial positions of acoustic emission (AE) sources in piping networks is essential for narrowing down degradation sites across pipe segments and weld joints. Traditional localization techniques often face limitations due to geometric complexity, wave dispersion, and signal detection issues resulting from incorrect time of arrival determination, sensor response differences, and variations in wave velocity. A regression-based mapping framework, in addition to conventional source localization, was introduced that leverages the geometric connectivity of the piping system to relate AE signal features to source-to-sensor distances and enable additional localization across multiple pipe segments and welds. The quadratic relationship between source-to-sensor distance and the amplitude-to-time-of-arrival ratio is demonstrated from the combined effects of geometric spreading and exponential material attenuation, which produce a nonlinear dependence well approximated by a second-order polynomial. The developed method is particularly effective when conventional linear localization or triangulation fails, such as when the AE source lies outside the sensor array. The approach was tested using the scaled advanced reactor vessel at the Mechanisms Engineering Test Loop (METL) facility at Argonne National Laboratory. Wave propagation behavior and acoustic connectivity were investigated experimentally and through numerical modeling. A finite element model was developed to simulate signal attenuation and time-of-arrival variations across different sensor configurations and was validated using pencil lead break tests. Across three representative pipe segments, the quadratic regression achieved with $${R}^{2}$$ R 2 values of 0.73–0.78, corresponding to localization errors less than 20% if the geometric mapping covers the source position in the regression model.

A two-stage acoustic source localization algorithm incorporating frequency-dependent atmospheric absorption

Underwater acoustics, governing the propagation and scattering of sound in the complex ocean environment characterized by multipath propagation, varied sound speed profile, and high ambient noise, is fundamental to acoustic applications like source localization, active sonar, and ocean acoustic tomography. Direction of Arrival (DOA) estimation is critical in this context, as it provides the direction location of acoustic sources using hydrophone arrays. Atomic norm minimization (ANM) theory has found its way into DOA estimation, alongside well-known compressive sensingbased and subspace-based methods, which enables gridless approaches to DOA estimation. This paper provides an overview of recent work on ANM-based DOA estimation. These new methods are motivated by techniques in atomic norm and the Vandermonde decomposition. Most of them have been proposed for locating the acoustic sources in challenging scenarios that require computational efficiency, high robustness performance, super-resolution capability, and even integration with deep learning techniques. These approaches provide important support for the further development of array signal processing in underwater acoustics.

Direction-of-arrival (DOA) estimation and localization of acoustic sources following major natural disasters, such as devastating earthquakes, are crucial for responding to immediate impacts and conducting search-and-rescue operations. With the rapid advancement of unmanned-aerial-vehicle (UAV) technologies, UAVs have become an excellent choice for carrying sensing and detection systems, as they offer better accessibility to disaster-stricken areas that are difficult for rescue teams to reach. However, a major challenge is that the acoustic sensing systems on UAVs are often affected by strong ego and environmental noise, leading to extremely low signal-to-noise ratios (SNRs), typically well below 0 dB, which makes most DOA estimation and source localization algorithms ineffective. To address this challenge, this work explores the design of dipping microphone arrays carried by UAVs and the associated DOA estimation algorithms. The major contributions are as threefold: 1) A dipping microphone array with a reconfigurable topology is designed, which significantly improves the SNR by adjusting the dipping length and enhances DOA estimation by configuring the array topology; 2) A maximum front-to-back ratio (MFBR) beamformer is developed to further mitigate the impact of UAV ego noise, further improving the SNR; 3) Building on the use of the dipping array and MFBR beamformer, a multiple-signal-classification (MUSIC) like algorithm is proposed to achieve accurate DOA estimation in challenging acoustic environments. Simulations and experiments are conducted to validate the effectiveness of the proposed design and algorithms.

A quasi-probabilistic brushing-based acoustic source localization technique for both burst and continuous sources

Microseismic/acoustic emission (AE) monitoring enables real-time, non-destructive observation of deformation and failure processes in rock during loading and unloading. Accordingly, this study designed two experimental schemes-sandstone loading and unloading-to comparatively investigate the spatiotemporal evolution characteristics of AE during sandstone failure under these distinct stress paths. Based on AE waveform time-frequency parameters and AE source location results obtained during testing, the failure evolution patterns of rock under both loading paths were analyzed. The results demonstrate that: (1) In both loading and load-unloading experiments, rock failure exhibited a distinct four-stage characteristic. Under pure loading conditions, failure concentrated near the point of catastrophic rupture, whereas unloading triggered premature rock fracturing, with a more pronounced AE response observed during the unloading phase. (2) For both loading paths, the dominant frequencies of AE waveforms were concentrated within the 0-200 kHz range. A distinct low-frequency (0-100 kHz), high-amplitude zone emerged prominently during Stage 4 in both cases. (3) AE source locations under load-unloading conditions revealed that during Stage 3-characterized by vertical loading combined with lateral unloading in the minimum principal stress direction-tensile failure cracks nucleated within the rock. Subsequently, during Stage 4 of the loading phase, these cracks propagated and coalesced, ultimately forming a macroscopic fracture surface on the sandstone specimen. (4) The AE source location results under pure loading failure conditions indicate that under uniaxial vertical loading, compression-shear failure fractures begin to develop within the rock mass during Stage 3. With continued loading in Stage 4, these shear fractures propagate through to the specimen surface, forming a through-going shear fracture plane.

Abstract To address the limitations of conventional acoustic emission (AE) source localization methods, particularly their neglect of refraction effects in layered media and poor convergence in iterative algorithms, this study proposes a Taylor first-order expansion hybrid optimization method based on the minimum travel time principle. The proposed approach innovatively integrates the minimum travel time principle with a layered velocity model to establish a control equation for the AE signal propagation path. Refraction points are rapidly determined using a projection-constrained dimensionality reduction technique, while a heuristic optimization strategy is employed to generate high-accuracy initial values. Iterative correction based on Taylor first-order expansion is then applied, forming a synergistic mechanism that ensures global convergence and local optimization. Experimental results demonstrate that, in two-layer and three-layer media models, the absolute localization error remains stable between 0.89 mm and 9.56 mm. Monte Carlo simulations further verify that the average error remains below 9.9 mm in five-layer media, with time-difference sensitivity remaining largely unaffected by the increasing number of layers. Notably, the hybrid optimization strategy achieves a 100% convergence rate, effectively overcoming the sensitivity of traditional Taylor-based methods to initial values. This study provides a theoretical breakthrough for precise AE source localization in layered media such as mine rock masses and composite structures, offering significant application potential in engineering nondestructive testing and geological disaster early warning systems.

Abstract To address the challenges posed by complex signal propagation paths in metallic discontinuous structures for impact localization, this study proposes a novel short-time Fourier transform (STFT)-based energy spectrum method combined with path planning and time-difference mapping for impact source localization in cylindrical metallic structures with hollow cavities. The method involves three core steps: first, perform cavity-aware meshing of discontinuous structural regions; second, the acoustic wave propagation paths of all gird points to four sensors is computed to generate theoretical time-of-arrival differences between all six sensor pairs; and finally, STFT-Energy Spectrum method is applied to compute the theoretical time-of-arrival differences, and absolute differences of gird points between theoretical and experimentally measured time delays are summed across all sensor pairs to form an error-matching matrix, with the grid point exhibiting the minimal total error identified as the impact source.Experimental validation across 228 impact scenarios demonstrated 1.11 cm average localization accuracy and 1.38% relative error, significantly outperforming conventional methods. This method provides a reliable solution for the health monitoring of discontinuous structures with holes.

A physics-informed machine learning (ML) framework for ocean acoustic source localization using matched field processing (MFP) is presented. A physics-informed neural network (PINN) predicts complex acoustic pressure fields from sparse pressure measurements and a known sound speed profile (SSP). These PINN-predicted replica fields are integrated into the MFP scheme, enabling fine-resolution source-receiver range estimation without requiring detailed geoacoustic bottom parameters. Validation with experimental data from the Shallow Water Evaluation Cell Experiment 1996 (SWellEx-96) demonstrates accurate range estimation, including in the challenging closest point of approach region. The method maintains performance when localizing from array element depths excluded during PINN training and under sparse-array configurations and moderate SSP mismatch. Compared to conventional model-based MFP, the method avoids full environmental characterization and mitigates environmental mismatch effects. Unlike purely data-driven ML methods, it incorporates the governing wave physics, producing physically consistent replicas and improving interpolation/extrapolation to ranges and array element depths that were not used in training. These results highlight the advantages of a physics-informed data-driven approach for ocean acoustic localization in realistic, data-limited environments.

Noise-robust acoustic emission source localization in reinforced concrete structures using a novel deep learning framework with skip connections

A new deep sparse unfolding network with variational Bayesian enhancement for high-resolution localization of acoustic sources

Several internal defect types can have an impact on structural performance and shorten its lifetime. Structural Health Monitoring (SHM) proposes a viable alternative by integrating a set of sensors for continuous structure monitoring. This enables early detection of the initiation and propagation of structural damage. Sensors permanent integration requires first determining their best placement, to ensure that a large area of the structure is monitored. However, using many sensors to cover a large area can have a negative impact on the structure's weight and thus, its performance. Hence, the main objective of this paper is to design an optimal sensor grid for acoustic source localization in plates using a network of four sensors along with triangulation algorithm. This work aims to validate experimentally the technique and to suggest a new procedure for implementing sensor networks for impact localization. The procedure is based on robust design methodology and sensor positions are determined based on the optimization of an objective function using the Taguchi SN ratio. A 400x400x2 mm aluminum plate is used herein, and the impact is generated by dropping a small steel ball at its center. Impact signals are acquired by piezoelectric sensors bonded to the plate's surface and captured by a four-channel oscilloscope. The efficiency of the proposed approach has been proved and the optimized sensor network located the impact with an error of 0.46 %.

Abstract The various nonlinear effects attributed to the propagation and the interaction of an acoustical shock with a ramp are investigated in this paper. The prevalent use of linear approximations in acoustic source localization techniques is a limitation to the accurate localization prediction required in military contexts. To address this issue, our approach seeks to identify different markers of nonlinearity within the acoustical shock wave framework, enlightening their reflective characteristics and the underlying physics. This study investigates the complex interaction between a high-amplitude acoustic pulse and a ramp, focusing on the reflection patterns of an acoustical shock. In particular, the single parameter used for the reflection pattern assessment is enhanced beyond its conventional formulation. The development of an irregular reflection detection algorithm is presented and serves as a fundamental component for a spectral analysis operating Fourier decomposition enabling a reflection-type classification solely based on time-signal measurements. This work contributes to the broader understanding of acoustic shock interactions and offers insights into improving the accuracy of source localization techniques, especially in situations where linear assumptions may prove to be limited.

Acoustic emission source localization in complex rock structures using sparse-grid fast path tracing method

To accurately characterize the non-radial motion of the source relative to the receiver, a three-dimensional (3D) model is essential. The extended Kalman filter (EKF) state matrix is employed to characterize the source's 3D motion. The measurement input for the EKF is the time delay between the direct and surface-reflected arrivals. The differences in the partial derivatives of the distance component have been identified and discussed. Through iterative filtering, a reliable estimate of the source's position in 3D space is obtained. Both simulations and experiments validate the effectiveness of the method, with experimental depth estimation errors within 1.5%.

For shallow underwater acoustic sources in deep-sea environments, the received signals in the direct wave zone exhibit distinct multipath arrival structures during large-depth reception, which can be exploited for passive source localization. This paper presents a method for estimating source depth and range through joint matching of arrival angles and time-delay differences between the direct wave and the first-order sea-surface reflected wave. The arrival angles are obtained from deep-sea particle velocity fields based on ray theory, while the time-delay differences are extracted from the autocorrelation function of the received signals. Simulation results show that both arrival angles and time-delay differences can be accurately estimated simultaneously at signal-to-noise ratios (SNRs) above 10 dB, allowing high-precision passive source localization with the proposed method. Under fixed source-receiver geometry, seabed topography and sea-surface roughness significantly alter acoustic propagation paths. Seabed variations affect paths of both direct and bottom-interacted waves, and sea-surface roughness specifically distorts the path of the first-order sea-surface reflected wave, thereby reducing the accuracy of time delay difference estimation in matched field processing. Validated using data from a 2020 deep-sea acoustic propagation experiment in the Western Pacific—where a single vector hydrophone captured signals from a towed source—the method effectively estimated the depth and range of shallow underwater acoustic sources within a 15 km range, with mean relative errors of 5.5% in depth and 6.2% in range. Compared to conventional methods that use only arrival angles or time-delay differences, this joint-matching strategy offers an efficient and robust solution for passive localization of underwater acoustic sources in the deep-sea direct wave zone.

Abstract Composite materials have become increasingly vital in modern industry due to their exceptional corrosion resistance and high strength-to-weight ratio. However, under varying environmental conditions and mechanical loading, these materials are prone to progressive damage and failure, making early detection and maintenance critical. Traditional acoustic emission (AE) analysis methods face significant limitations in damage monitoring due to the anisotropic nature of composites, which leads to severe modal dispersion in acquired signals and complicates mode identification. These challenges undermine the accuracy and reliability of conventional source localization techniques.To address these issues, this study proposes a novel AE source localization method that combines modal arrival time and wave velocity analysis. The approach leverages time difference of arrival (TDOA) and modal wave velocity data captured by a three-sensor array to determine both the angular direction and precise planar coordinates of AE sources. Experimental validation demonstrates that the proposed method achieves high localization accuracy within a 350 mm radius from the sensor array center. These findings advance the capability for real-time structural health monitoring in composite structures, offering improved diagnostic precision for industrial applications.