Time Difference Of Arrival Research Articles

Performance of recurrent neural networks with Monte Carlo dropout for predicting pharmacokinetic parameters from dynamic contrast‐enhanced magnetic resonance imaging data

AbstractPurposeTo quantitatively evaluate the performance of two types of recurrent neural networks (RNNs), long short‐term memory (LSTM) and gated recurrent units (GRU), using Monte Carlo dropout (MCD) to predict pharmacokinetic (PK) parameters from dynamic contrast‐enhanced magnetic resonance imaging (DCE‐MRI) data.MethodsDCE‐MRI data for simulation studies were synthesized using the extended Tofts model and a population‐averaged arterial input function (AIF). The ranges of PK parameters for training the RNNs were determined from data of patients with brain tumors. The effects of the number of training samples, number of hidden units, dropout rate (DR), and bolus arrival time delay and dispersion in AIF on the accuracy of the PK parameters were investigated, and the uncertainties for different DRs and peak signal‐to‐noise ratios (PSNRs) were quantified. For comparison, PK parameters were estimated using the nonlinear least‐squares method. In the clinical studies, the PK parameter and uncertainty images were generated by applying the trained RNNs to DCE‐MRI data.ResultsCompared with GRU, the computational cost for training the LSTM was significantly higher. The prediction accuracy of GRU decreased with decreasing numbers of training samples and hidden units, whereas the performance of LSTM remained stable. Despite an increased computational cost, MCD reduced the prediction error at low PSNR and improved the quality of PK parameter images. The simulation results recommended using a DR of 0.25–0.5 at low PSNR and ≤ 0.25 for other PSNRs. The clinical studies recommended using a DR of 0.25 and 0.5 for LSTM and GRU, respectively.ConclusionsMCD is effective in quantifying uncertainty in PK parameter prediction from DCE‐MRI data and improves their performance, particularly at low PSNR; however, at the expense of increased computational cost. This study helps deepen our understanding of RNNs with MCD and select suitable hyperparameters for creating an RNN architecture for DCE‐MRI studies.

A thunderstorm climatology of Romania (1941―2022)

Thunderstorms and their associated hazards pose a significant threat to society and the economy. In Europe, between 1980 and 2022, thunderstorms caused an estimated EUR 190 billion in economic losses. In Romania, the rate of cloud-to- ground (CG) lightning fatality is one of the highest in Europe. This study aims to reevaluate the risk of lightning in Romania by updating the CG lightning climatol- ogy for the period 2010–2022 and analysing long-term changes (1941-2022) in the frequency of thunderstorm environments. Data were obtained from the Arrival Time Difference Lightning Network (ATDnet) and the ERA5 reanalysis. On average, ap- proximately 760,000 flashes occurred each year in Romania during the study period, with a maximum mean annual flash density of approximately 6.5 flashes km−2 yr−1 over the Southern Carpathians. Approximately 75% of all flashes were detected in June–August, with a peak in the afternoon hours. There is a general decrease in flash density in Romania, but this trend is statistically significant only in western Romania. The highest number of hours with thunderstorm environments (270–390 hours yr−1) was predominantly observed in the mountainous regions of central and western Ro- mania. There is a significant increase in the number of hours with thunderstorms per decade, especially in eastern and southern Romania (>4 hours per decade). The re- sults of this study allow for a better understanding of the risk posed by CG lightning in Romania, improved forecasting of thunderstorms, and serve as a basis for developing mitigation strategies to reduce the impact of CG lightning.

Accuracy and Precision of Earthquake Location Programs: Insights from a Synthetic Controlled Experiment

Abstract Earthquake location programs employ diverse approaches to address the challenges posed by incomplete knowledge and simplified representation of complex Earth structures. Assessing their reliability in location and uncertainty characterization remains challenging as benchmark datasets with known event locations are rare, and usually confined to particular sources, such as quarry blasts. This study evaluates eight earthquake location methods (GrowClust, HypoDD, Hypoinverse, HypoSVI, NonLinLoc, NonLinLoc_SSST, VELEST, and XCORLOC) through a controlled synthetic computational experiment on 1000 clustered earthquakes based on the setting of the 2019 Ridgecrest, California, earthquake sequence. We construct a travel-time dataset using the fast-marching method and a 3D velocity model extracted from the Community Velocity Model, supplemented with a von Karman perturbation to represent small-scale heterogeneity, and including elevation effects. Picking errors, phase availability, and outliers are introduced to mimic difficulties encountered in seismic network monitoring. We compare the location results from eight programs applied to the same travel-time dataset and 1D velocity structure against the ground-truth locations. For this aftershock sequence, our results reveal the superior accuracy and precision of differential time-based location methods compared to single-event location methods. The results validate the significance of compensating for deviations from assumed 1D velocity structure either by path or site correction modeling or by cancellation of unmodeled structure using differential arrival times. We also evaluate the uncertainty quantification of each program and find that most of the programs underestimate the errors.

Unreported large errors in a common method for sound source localization of marine mammals.

Confidence intervals of location of calling marine mammals, derived from time differences of arrival (TDOA) between receivers, depend on errors of TDOAs, receiver location, clocks, sound speeds, and location method. Simulations demonstrate Ishmael, a TDOA locator based on uncorrected least squares minimization (ULSM), yields errors with mean, standard deviation, and maximum of 0.1, 0.2, and 0.9 km, respectively, due to sensitivity to inputs and numerical implementation when applied to scenarios with minuscule errors; e.g., five clock-synchronized receivers residing on the vertices of a square with one in its center. This sensitivity can mask other causes of location error due to small uncertainties in receiver location and sound speed. Realistic uncertainties of sound speed up to ±7.5 m/s lead to errors up to 4 km. With unsynchronized clocks and common practice of correcting TDOA from synchronization measurements at the start and end of an experiment, errors of location are 10 to 1000 km. These problems occur because ULSM was not designed to account for all errors. ULSM is also available in PAMGuard and other systems and is used to study behavior and abundance of calling marine mammals. ULSM is briefly compared to another method designed to account for errors.

Deep Learning-Aided Acoustic Source Localization in Thin-Walled Waveguides

The timely detection of fault locations represents a critical task in Structural Health Monitoring (SHM) of thinwalled elements. In particular, the localization of acoustic emission sources is particularly important for the identification of damages caused by stress and can be achieved by estimating the Difference in Time of Arrival (DToA) between the waves captured by a sparse sensor array. In this work, a novel method for DToA extraction suitable for isotropic structures is proposed. Our approach is based on the combination of Convolutional Neural Networks (CNNs) and a dispersion compensation operator, the Warped Frequency Transform (WFT). CNNs are deployed to enhance the localization process against the detrimental losses caused by non-ideal conditions, such as the presence of reflections and multiple propagation modes. The results show that such method yields localization errors of a few centimetres, with an average below 2 cm, when only relying on hundreds of real data for training.

Method of missile formation positioning and tracking based on TDOA-DOA

Abstract With the development of science and technology, the battlefield environment of modern warfare has become complicated and complex, and it has become an important task to obtain information about enemy targets quickly and accurately. Accordingly, researchers from various countries have explored various positioning methods, which can be divided into active and passive positioning, among which passive positioning has received more attention due to its excellent concealment and anti-jamming ability. In this paper, the two methods of passive positioning, Direction of Arrival (DOA), and Time Difference of Arrival (TDOA), are used to study the positioning of missile formations in two- and three-dimensional space and to analyse the factors affecting their positioning accuracy.

High-accuracy detection and location of gas leaks based on adaptive weighted data fusion using an ultrasonic transducer array

This paper proposes a gas leak detection and location method with high accuracy based on an adaptive weighted data fusion algorithm using an ultrasonic transducer array. First, the approximate entropy theory is combined with a support vector machine (SVM) to judge whether there is a gas leak, achieving a recognition accuracy of 95.69%. Once a leak is identified, a leak localisation algorithm is applied. To obtain more stable features and reduce environmental noise, the sound pressure amplitude and time-delay difference are extracted using an optimised signal processing method and normalised least mean squares (NLMS) adaptive filter, respectively. The energy decay (ED) localisation algorithm is then fused with the time difference of arrival (TDOA) algorithm to determine the location of the leak. Experimental results demonstrate that under a pressure of 15 kPa in the measured container with a distance of 1 m between the leak and transducer array, the location error is within 5 mm, which is more accurate than previously proposed methods. Additionally, the adaptive weighted data fusion algorithm overcomes the limitations of the TDOA and ED algorithms.

An Experimental Testbed for the Performance Evaluation of Optical Time Difference of Arrival Based Indoor Positioning

An Experimental Testbed for the Performance Evaluation of Optical Time Difference of Arrival Based Indoor Positioning

WSN Localization in Smart Cities Using Hybrid (TDOA & RSSI) Localization Techniques with Extended Kalman Filter

In the context of smart cities, localization technologies are essential for tracking objects in both indoor and outdoor environments. The characteristics of urban settings such as building density, signal interference, and multipath propagation significantly impact the accuracy of localization algorithms. This paper reviews existing localization techniques, categorizing them into range-based and range-free methods, and discusses key algorithms including trilateration, multilateration, and triangulation. We propose a hybrid localization framework that combines Time Difference of Arrival (TDOA) and Received Signal Strength Indicator (RSSI) techniques, enhanced by an Extended Kalman Filter (EKF) for improved accuracy and robustness. Additionally, we present a design architecture for a transmitter and receiver system utilizing Long Range (LoRa) technology, facilitating the development of low-cost, low-power tracking devices that operate independently of Global Positioning System (GPS). Our approach aims to enhance the efficacy of localization in smart city applications, ultimately contributing to improved urban management and safety.

Adaptive Forecasting of Nuclear Power Plant Operational Status Under Sensor Concept Drift Using a Bridging Distribution Adaptive Network.

A large number of sensors are required to collect information during the operation of nuclear power plants to ensure their absolutely safe operation. However, because of the unique nature of nuclear reactions, the physical environment of nuclear power production is prone to changes, leading to concept drift in the data collected by the sensors. Concept drift describes the phenomenon of sample distribution changing over time, which typically negatively impacts the model's training and inference processes. We found that nongradual distribution changes could be guided by generating transitional intermediary distributions within the distribution, thereby achieving a gradual change process. Based on this, we designed a bridging distribution adaptive network (BDAN), which consisted of identical-depth TDoA (time difference of arrival) homomorphic backbone neural networks on both sides with a latent adaptive bridging module in the middle. By calculating the distribution differences over multiple timesteps, a series of bridge distributions were generated to guide the gradients in the latent space, updating the parameters of the latent adaptive guiding module in a directional manner and enabling the model to perceive nongradual distribution changes in the time domain. Experimental results showed that the BDAN outperformed the previous state-of-the-art benchmark methods by 5.6% in terms of mean squared error in the nuclear power data prediction task under concept drift, achieving the best fault prediction performance.

Distributed Passive Positioning and Sorting Method for Multi-Network Frequency-Hopping Time Division Multiple Access Signals.

When there are time division multiple access (TDMA) signals with large bandwidth, waveform aliasing, and fast frequency-hopping in space, current methods have difficulty achieving the accurate localization of radiation sources and signal-sorting from multiple network stations. To solve the above problems, a distributed passive positioning and network stations sorting method for broadband frequency-hopping signals based on two-level parameter estimation and joint clustering is proposed in this paper. Firstly, a two-stage filtering structure is designed to achieve control filtering for each frequency point. After narrowing down the parameter estimation range through adaptive threshold detection, the time difference of arrival (TDOA) and the velocity difference of arrival (VDOA) can be obtained via coherent accumulating based on the cross ambiguity function (CAF). Then, a multi-station positioning method based on the TDOA/VDOA is used to estimate the position of the target. Finally, the distributed joint eigenvectors of the multi-stations are constructed, and the signals belonging to different network stations are effectively classified using the improved K-means method. Numerical simulations indicate that the proposed method has a better positioning and sorting effect in low signal-to-noise (SNR) and low snapshot conditions compared with current methods.

Single Station Seismic Event Location With a Neural‐Guided Mixture Model: Model‐Based Weighting for Improved Accuracy

AbstractIn many regions of the world with complex seismic activity, the availability of seismic stations is limited. This motivated us to develop a probabilistic method for single‐station earthquake location. Single station location methods may provide an efficient and cost‐effective way to monitor small‐magnitude seismic events over large areas of the world. This method relies on the outputs of two neural networks, which predict the initial spatial location of specific areas (convolutional) used in the prior probability and seismic phase recognition (combination of convolutional and Long Short Term Memory) used in the likelihood function. These estimates and a 3D travel time likelihood function are used in a Bayesian framework. In this model, the initial spatial location is expressed as a Multivariate Gaussian Mixture Model, and the likelihood function refines the localization using the arrival time difference between S and P waves in a 3D space. The likelihood function uses a Fast Marching Eikonal method with a 3D velocity structure. We tested this method on the Ridgecrest 2019 seismic sequence (Mw = 7.1) and on West Texas and found promising results for locating with a single station. Our approach demonstrates that using a Bayesian model with strong priors obtained through deep learning algorithms is effective. Our model can predict epicenter and depth with mean errors of 7.55 km in longitude, 13.54 km in latitude and 0.630 km in depth at the two different locations studied. This method demonstrates the effectiveness of combining deep learning for initial estimations with probabilistic 3D location refinement.

A Seismic Precursor 15 min Before the Giant Eruption of Hunga Tonga‐Hunga Ha'apai Volcano on 15 January 2022

AbstractThe 15 January 2022, eruption at Hunga Tonga‐Hunga Ha'apai (HTHH) volcano started shortly after 4:00UTC. There had been noted unconfirmed precursory events. We analyzed seismometer data recorded in Fiji and Futuna, the closest stations operated during the eruption and located over 750 km away. We extracted Rayleigh waves and estimated their powers and source directions, assuming retrograde particle motions. We found a Rayleigh wave from the HTHH's direction about 15 min before the eruption onset. The arrival time difference of the Rayleigh wave between the two stations was consistent with that of the M5.8 earthquake during the eruption located beneath the HTHH. Referring to other seismic signals and satellite images, we concluded that the Rayleigh wave was the most significant eruption precursor with no apparent surface activity. Including our findings and results of previous studies, we propose a scenario of the beginning of the caldera‐forming eruption.

Numerical investigation of bandgap characteristics in integrated metallic metafilters for nonlinear guided wave applications

Guided waves propagating in nonlinear media, featuring second harmonic generation, represent a promising avenue for early-stage damage detection due to their high sensitivity and long-range propagation capabilities. However, nonlinear ultrasonic measurements are hindered by nonlinearities induced by the experimental system, necessitating careful calibrations that have restricted their application to laboratory settings. While several phononic crystal and metamaterial designs have been devised to enhance nonlinear-based ultrasonic testing, most are tailored for suppressing second harmonics within a frequency range of 100–300 kHz, primarily utilizing low-frequency excitation. In this paper, we propose a metallic ring-shaped metafilter designed to explore high-order bandgaps. To fully understand the bandgap characteristics, we begin by analyzing mode shapes, providing insights into the underlying wave mechanics. The efficacy of the designed filter is subsequently assessed through 3D time step elastodynamic simulations. In addition, this study underscores the significance of parameters such as the number of rings employed in the filter, signal duration, and bandgap width in optimizing its performance. Furthermore, the observed mode conversion phenomena from S0 to A0 guided wave modes underscore the filter’s capacity to influence guided wave propagation. The defect localization technique, based on the time difference of arrival of second-order wave modes, accurately predicts the defect location with an error margin of less than 0.2%. The present investigation showcases advancements in the sensitivity of nonlinear-based guided wave testing for characterizing microstructural changes, promising substantial potential for detecting incipient damage in practical structural health monitoring applications.

A theoretical analysis of the logging-while-drilling dipole acoustic reflection measurement

Post-drilling wireline acoustic single-well imaging technology can now detect geological structures tens of meters away from boreholes. Further development of this single-well imaging technology in the logging-while-drilling (LWD) environment will have significant values in real-time applications such as geosteering and reservoir navigation. Based on the wireline imaging application, we propose a new method for the LWD application. In wireline imaging, the four-component (4C) dipole acoustic data are azimuthally rotated to scan the reflectors around the borehole. In LWD, azimuthal scanning is achieved by drilling rotation such that the 4C dipole system in the wireline is replaced by a one-dipole-source and two-receiver LWD system, where the two receivers are mounted on opposite sides of the drill collar. For the LWD application, we first developed the theory for LWD dipole shear-wave reflection imaging and validated the theory using 3D finite-difference waveform modeling. Using the analytical solution, we analyzed the far-field radiation directivity of an acoustic LWD dipole source and the effect of drilling rotation on the shear-wave reflection imaging using the LWD acoustic system. The LWD analysis results show that, for fast formations, the SH-wave is the dominant component for imaging, whereas for slow formations, the P-wave becomes important and can be used for imaging. Our results also indicate that the reflection data acquired by the system are affected by the speed of drilling rotation. The take-off azimuth at the wave radiation may be different from the incident azimuth at the wave reception. Knowing the rotation speed, this azimuth difference can be corrected. A further advantage of using the oppositely mounted receivers is that the reflected wave arrives earlier (later) at the front (back)-side receiver; thus, the arrival time difference between the receivers can be used to eliminate the 180°-azimuth ambiguity of dipole acoustic imaging. For reflection imaging, using the proposed LWD system configuration, we tested its azimuth sensitivity and validated its 180°-ambiguity solution using synthetic LWD and field wireline dipole data. The results of this work, therefore, provide a theoretical foundation for the development of the LWD acoustic reflection imaging system.

A triangular acoustic emission source location method considering its anisotropic propagation behavior on the surface of finger-joint-laminated board

ABSTRACT To understand the anisotropic propagation characteristics of acoustic emission (AE) signals in finger-joint-laminated boards, an improved planar localization method for AE sources is proposed. Initially, the AE source was generated by the pencil-lead break (PLB) at a fixed position on the surface of the glued wood specimen, and two AE sensors were arranged in different directions at a fixed distance of 60 mm, and the AE propagation velocity was calculated according to the time difference of arrival (TDOA). For clarity, the grain parallel direction was defined as 0 degrees, with velocities measured every 10 degrees through a counterclockwise rotation. Using these measurements, two angular anisotropic wave velocity models were developed. Subsequently, three AE sensors arranged linearly on the specimen surface pinpointed the AE source’s location. This source location problem was formulated as a multi-parameter optimization model, constrained by the geometric relationships between the AE source and the sensors. A particle swarm optimization (PSO) algorithm was utilized to estimate the AE source’s angles relative to each sensor. The findings indicate that the located AE sources deviated by 6.0–19.0 mm from the predetermined PLB positions, with inaccuracies largely attributed to the anisotropic AE wave velocity models’ precision.

Numerical modelling of bridge deck reinforcement corrosion based on analysis of GPR data

AbstractThis study explores the impact of corrosion on Ground Penetrating Radar (GPR) responses through practical experiments and numerical modelling, focusing on rebar diameter reduction, corrosion product layer thickness, crack formation and corrosion product filling in vertical and transverse crack. Practical experiments involved GPR testing of reinforced concrete slab. By analyzing B-scans we identify areas with moderate and severe corrosion. Numerical modelling using the Finite Difference Time Domain (FDTD) Method to model GPR signal propagation in a concrete bridge deck with corrosion is applied. Key finding includes a significant 26.70% increase in reflected wave amplitude when corrosion product filling in vertical crack increased by 400%, highlighting its extensive effect on signal GPR propagation. Reduced rebar diameter led to a 9.79% amplitude decrease and a 0.06 ns arrival time delay. Increased corrosion product layer thickness primarily affected arrival time with a 0.06 ns extension but significantly amplified GPR signal amplitude. These findings offer insights for improving GPR based corrosion detection and assessment methods, leading to more robust systems for concrete bridge deck inspection and maintenance. This paper contributes to understanding how corrosion affects the signal that is detected by GPR. This information can be used to improve the way that we manage and assess corrosion in concrete bridge deck.

Retraction notice: Enhanced gray wolf optimization for estimation of time difference of arrival in WSNs

Retraction notice: Enhanced gray wolf optimization for estimation of time difference of arrival in WSNs

Quasi-real-time earthquake relocation and monitoring in the northeastern Noto Peninsula

The seismicity rate markedly increased in the northeastern Noto Peninsula of Ishikawa Prefecture around the end of 2020, with an Mw6.2 event on 5 May, 2023, followed by many aftershocks. Previous earthquake relocation studies have detected upward migration of microearthquakes via multiple faults and clusters, suggesting the involvement of crustal fluids in this sequence. Since some active faults exist near the source region, there was concern that this sequence could lead to a larger earthquake; it became a reality with the Mw7.5 earthquake on 1 January 2024. This study aims to develop an algorithm to precisely relocate microearthquake hypocenters in quasi-real time for better monitoring. A fine view of seismicity requires relative relocation methods such as the double-difference (DD) method with numerous arrival time difference data precisely measured by the waveform correlation analysis. However, the standard DD method has the disadvantage of huge computational costs when data increase, making it unsuitable for real-time monitoring in such situations. We developed a quasi-real-time algorithm that relocates only new earthquakes using the DD method each time a new time window of data is added. The major improvement is that our method incorporates a traditional simple relative relocation and preserves constraints between different time windows; the relative locations of new events are constrained from reference events that were already relocated in the previous time windows. We tested a daily relocation algorithm on 11,546 events from 19 June, 2022, to 31 May, 2023, in the Noto Peninsula earthquake sequence. We found that our modification substantially reduced artificial hypocenter offsets between different time windows and succeeded in resolving the fine fault structures from the cloud-like distribution of initial hypocenters. If we do not impose constraints between different windows, the relocated hypocenters are scattered and do not show fine planar structures. Our algorithm greatly reduces the computational cost, allowing for quasi-real-time earthquake relocation and monitoring. We hope this algorithm will help monitor the spatio-temporal distribution of future earthquake sequences.Graphical

Enhanced Method for Localization of Partial Discharges in Oil-Filled Transformers Using Acoustic Emission Signals

The accurate measurement of partial discharges (PD) in power transformers is a critical component in identifying faults and planning effective maintenance strategies. Acoustic emission (AE) sensing technology is suitable for detecting partial discharges in transformers because it is less affected by external electromagnetic interference and detects PD non-invasively. This paper delves into examining the relationship between the detection intensity, the specific type of PD source, and the distance to the discharge source, using AE sensors. It was observed that AE wave intensities from corona discharges tend to be relatively strong, indicating a higher detection probability. Creepage discharges usually exhibit the next level of intensity, followed by the PD occurring in bubbles. It is considered that the intensity of the AE waves varies significantly depending on both the speed at which the discharge propagates and the medium and volume of the discharge space. Furthermore, this study has conducted experimental comparisons among three distinct methods for calculating the Time Difference of Arrival (TDOA) in localization calculations. These methods are the energy criterion, Generalized Cross-Correlation (GCC), and GCC with Phase Transformation (PHAT). The experimental results suggest that the energy criterion method is particularly effective when sensors are distributed placement around the entire tank. In contrast, the GCC-PHAT method shows greater suitability in scenarios where sensors are centralized placement in certain sections of the tank. It was clearly observed that the GCC-PHAT method, which suppresses the impact of noise and reflections, consistently achieves higher estimation accuracy in comparison to the standard GCC method. Notably, this method has shown the ability to maintain its accuracy levels even in cases of low discharge intensities. The implications of these findings are significant, as they could improve the precision and effectiveness of maintenance diagnostics in power transformers.