Deep Learning Based Inversion of Locally Anisotropic Weld Properties from Ultrasonic Array Data

The ability to reliably detect and characterise defects embedded in austenitic steel welds depends on prior knowledge of microstructural descriptors, such as the orientations of the weld’s locally anisotropic grain structure. These orientations are usually unknown but it has been shown recently that they can be estimated from ultrasonic scattered wave data. However, conventional algorithms used for solving this inverse problem incur a significant computational cost. In this paper, we propose a framework which uses deep neural networks (DNNs) to reconstruct crystallographic orientations in a welded material from ultrasonic travel time data, in real-time. Acquiring the large amount of training data required for DNNs experimentally is practically infeasible for this problem, therefore a model based training approach is investigated instead, where a simple and efficient analytical method for modelling ultrasonic wave travel times through given weld geometries is implemented. The proposed method is validated by testing the trained networks on data arising from sophisticated finite element simulations of wave propagation through weld microstructures. The trained deep neural network predicts grain orientations to within 3° and in near real-time (0.04 s), presenting a significant step towards realising real-time, accurate characterisation of weld microstructures from ultrasonic non-destructive measurements. The subsequent improvement in defect imaging is then demonstrated via use of the DNN predicted crystallographic orientations to correct the delay laws on which the total focusing method imaging algorithm is based. An improvement of up to 5.3 dB in the signal-to-noise ratio is achieved.

Download Full-text

Real-time super-resolution mapping of locally anisotropic grain orientations for ultrasonic non-destructive evaluation of crystalline material

Neural Computing and Applications ◽

10.1007/s00521-021-06670-8 ◽

2021 ◽

Author(s):

Jonathan Singh ◽

Katherine Tant ◽

Andrew Curtis ◽

Anthony Mulholland

Keyword(s):

Real Time ◽

Computational Cost ◽

Super Resolution ◽

Structural Similarity ◽

Image Resolution ◽

Training Data ◽

Generative Adversarial Networks ◽

Desktop Computer ◽

Data Set ◽

Grain Orientations

AbstractEstimating the spatially varying microstructures of heterogeneous and locally anisotropic media non-destructively is necessary for the accurate detection of flaws and reliable monitoring of manufacturing processes. Conventional algorithms used for solving this inverse problem come with significant computational cost, particularly in the case of high-dimensional, nonlinear tomographic problems, and are thus not suitable for near-real-time applications. In this paper, for the first time, we propose a framework which uses deep neural networks (DNNs) with full aperture, pitch-catch and pulse-echo transducer configurations, to reconstruct material maps of crystallographic orientation. We also present the first application of generative adversarial networks (GANs) to achieve super-resolution of ultrasonic tomographic images, providing a factor-four increase in image resolution and up to a 50% increase in structural similarity. The importance of including appropriate prior knowledge in the GAN training data set to increase inversion accuracy is demonstrated: known information about the material’s structure should be represented in the training data. We show that after a computationally expensive training process, the DNNs and GANs can be used in less than 1 second (0.9 s on a standard desktop computer) to provide a high-resolution map of the material’s grain orientations, addressing the challenge of significant computational cost faced by conventional tomography algorithms.

Download Full-text

Tracking and Visualizing Signs of Degradation for Early Failure Prediction of Rolling Bearings

Journal of Robotics and Mechatronics ◽

10.20965/jrm.2021.p0629 ◽

2021 ◽

Vol 33 (3) ◽

pp. 629-642

Author(s):

Sana Talmoudi ◽

Tetsuya Kanada ◽

Yasuhisa Hirata ◽

◽

Keyword(s):

Real Time ◽

Training Data ◽

High Dimensional ◽

Time Data ◽

Early Failure ◽

Rolling Bearings ◽

Full Spectrum ◽

Time Frequency ◽

Vibration Data ◽

Visualization Technology

Predictive maintenance, which means detection of failure ahead of time, is one of the pillars of Industry 4.0. An effective method for this technique is to track early signs of degradation before failure occurs. This paper presents an innovative failure predictive scheme for machines. The proposed scheme combines the use of the full spectrum of vibration data from the machines and a data visualization technology. This scheme requires no training data and can be started quickly after installation. First, we proposed to use the full spectrum (as high-dimensional data vectors) with no cropping and no complex feature extraction and to visualize the data behavior by mapping the high-dimensional vectors into a two-dimensional (2D) map. This ensures simplicity of the process and less possibility of overlooking important information as well as provide a human-friendly and human-understandable output. Second, we developed a real-time data tracker that can predict failure at an appropriate time with sufficient allowance for maintenance by plotting real-time frequency spectrum data of the target machine on a 2D map created from normal data. Finally, we verified our proposal using vibration data of bearings from real-world test-to-failure measurements obtained from the IMS dataset.

Download Full-text

Wind Turbine Noise Reduction from Seismological Data

10.5194/egusphere-egu21-4907 ◽

2021 ◽

Author(s):

Janis Heuel ◽

Wolfgang Friederich

Keyword(s):

Signal To Noise Ratio ◽

Low Cost ◽

Computational Cost ◽

Noise Signal ◽

Noise Model ◽

Threshold Function ◽

High Signal ◽

Time Data ◽

Negative Effects ◽

Before And After

Over the last years, installations of wind turbines (WTs) increased worldwide. Owing to negative effects on humans, WTs are often installed in areas with low population density. Because of low anthropogenic noise, these areas are also well suited for sites of seismological stations. As a consequence, WTs are often installed in the same areas as seismological stations. By comparing the noise in recorded data before and after installation of WTs, seismologists noticed a substantial worsening of station quality leading to conflicts between the operators of WTs and earthquake services.In this study, we compare different techniques to reduce or eliminate the disturbing signal from WTs at seismological stations. For this purpose, we selected a seismological station that shows a significant correlation between the power spectral density and the hourly windspeed measurements. Usually, spectral filtering is used to suppress noise in seismic data processing. However, this approach is not effective when noise and signal have overlapping frequency bands which is the case for WT noise. As a first method, we applied the continuous wavelet transform (CWT) on our data to obtain a time-scale representation. From this representation, we estimated a noise threshold function (Langston & Mousavi, 2019) either from noise before the theoretical P-arrival (pre-noise) or using a noise signal from the past with similar ground velocity conditions at the surrounding WTs. Therefore, we installed low cost seismometers at the surrounding WTs to find similar signals at each WT. From these similar signals, we obtain a noise model at the seismological station, which is used to estimate the threshold function. As a second method, we used a denoising autoencoder (DAE) that learns mapping functions to distinguish between noise and signal (Zhu et al., 2019).In our tests, the threshold function performs well when the event is visible in the raw or spectral filtered data, but it fails when WT noise dominates and the event is hidden. In these cases, the DAE removes the WT noise from the data. However, the DAE must be trained with typical noise samples and high signal-to-noise ratio events to distinguish between signal and interfering noise. Using the threshold function and pre-noise can be applied immediately on real-time data and has a low computational cost. Using a noise model from our prerecorded database at the seismological station does not improve the result and it is more time consuming to find similar ground velocity conditions at the surrounding WTs.

Download Full-text

The Influence of Crystallographic Orientations of Grains on Microstructurally Small Cracks Using Crystal Plasticity and Random Grain Structure

Volume 6: Materials and Fabrication ◽

10.1115/pvp2006-icpvt-11-93363 ◽

2006 ◽

Cited By ~ 1

Author(s):

I. Simonovski ◽

L. Cizelj

Keyword(s):

Crystal Plasticity ◽

Crack Tip ◽

Maximum Load ◽

Grain Structure ◽

Crystal Plasticity Model ◽

Grain Orientations ◽

Small Cracks ◽

Crack Tip Opening ◽

Soft Grains ◽

Crystallographic Orientations

A plane-strain finite element crystal plasticity model of microstructurally small stationary crack emanating at a surface grain in a 316L stainless steel is proposed. The model consisting of 212 randomly shaped, sized and oriented grains is loaded monotonically in uniaxial tension to a maximum load of 1.12Rp0.2 (280 MPa). The influence that a random grain structure imposes on a Stage I crack is assessed by calculating the crack tip opening (CTOD) and sliding displacements (CTSD), considering also different crystallographic orientations. It is shown that certain crystallographic orientations result in a cluster of soft grains around the crack-containing grain. In these cases the crack tip can become apart of the localized strain, resulting in a large CTOD value. This effect, resulting from the overall grain orientations and sizes, can have a greater impact on the CTOD than the local grain orientation. On the other hand, when a localized soft response is formed away from the crack, the localized strain does not affect the crack tip directly, resulting in a small CTOD value. The resulting difference in CTOD can be up to a factor of 4, depending upon the crystallographic set. Grains as far as 6xCracklength significantly influence that crack tip parameters. It was also found the a larger crack-containing grain tends to increase the CTOD.

Download Full-text

A Comprehensive Comparative Analysis of the Basic Theory of the Short Term Bus Passenger Flow Prediction

Symmetry ◽

10.3390/sym10090369 ◽

2018 ◽

Vol 10 (9) ◽

pp. 369 ◽

Cited By ~ 4

Author(s):

Huawei Zhai ◽

Licheng Cui ◽

Yu Nie ◽

Xiaowei Xu ◽

Weishi Zhang

Keyword(s):

Real Time ◽

Public Transportation ◽

Transportation Systems ◽

Training Data ◽

Time Data ◽

Short Term ◽

Passenger Flow ◽

Modeling Process ◽

Automatic Passenger Counters ◽

Public Transportation Systems

In order to meet the real-time public travel demands, the bus operators need to adjust the timetables in time. Therefore, it is necessary to predict the variations of the short-term passenger flow. Under the help of the advanced public transportation systems, a large amount of real-time data about passenger flow is collected from the automatic passenger counters, automatic fare collection systems, etc. Using these data, different kinds of methods are proposed to predict future variations of the short-term bus passenger flow. Based on the properties and background knowledge, these methods are classified into three categories: linear, nonlinear and combined methods. Their performances are evaluated in detail in the major aspects of the prediction accuracy, the complexity of training data structure and modeling process. For comparison, some long-term prediction methods are also analyzed simply. At last, it points that, with the help of automatic technology, a large amount of data about passenger flow will be collected, and using the big data technology to speed up the data preprocessing and modeling process may be one of the directions worthy of study in the future.

Download Full-text

Nature-Inspired Drone Swarming for Real-Time Aerial Data-Collection Under Dynamic Operational Constraints

Drones ◽

10.3390/drones3030071 ◽

2019 ◽

Vol 3 (3) ◽

pp. 71 ◽

Cited By ~ 4

Author(s):

Hanno Hildmann ◽

Ernö Kovacs ◽

Fabrice Saffre ◽

A. F. Isakovic

Keyword(s):

Data Collection ◽

Real Time ◽

Precision Agriculture ◽

Computational Cost ◽

Civil Defense ◽

Time Data ◽

Communication Infrastructure ◽

Distributed Approach ◽

Nature Inspired Algorithm ◽

Real Time Data

Unmanned Aerial Vehicles (UAVs) with acceptable performance are becoming commercially available at an affordable cost. Due to this, the use of drones for real-time data collection is becoming common practice by individual practitioners in the areas of e.g., precision agriculture and civil defense such as fire fighting. At the same time, as UAVs become a house-hold item, a plethora of issues—which can no longer be ignored and considered niche problems—are coming of age. These range from legal and ethical questions to technical matters such as how to implement and operate a communication infrastructure to maintain control over deployed devices. With these issues being addressed, approaches that focus on enabling collectives of devices to operate semi-autonomously are also increasing in relevance. In this article we present a nature-inspired algorithm that enables a UAV-swarm to operate as a collective which provides real-time data such as video footage. The collective is able to autonomously adapt to changing resolution requirements for specific locations within the area under surveillance. Our distributed approach significantly reduces the requirements on the communication infrastructure and mitigates the computational cost otherwise incurred. In addition, if the UAVs themselves were to be equipped with even rudimentary data-analysis capabilities, the swarm could react in real-time to the data it generates and self-regulate which locations within its operational area it focuses on. The approach was tested in a swarm of 25 UAVs; we present out preliminary performance evaluation.

Download Full-text

Predicting On-axis Rotorcraft Dynamic Responses Using Machine Learning Techniques

Journal of the American Helicopter Society ◽

10.4050/jahs.65.032004 ◽

2020 ◽

Vol 65 (3) ◽

pp. 1-12

Author(s):

Ryan D. Jackson ◽

Michael Jump ◽

Peter L. Green

Keyword(s):

Machine Learning ◽

Real Time ◽

Computational Cost ◽

Response Term ◽

Dynamic Responses ◽

Training Data ◽

Machine Learning Techniques ◽

Human In The Loop ◽

Roll Rate ◽

Gp Model

Physical-law-based models are widely utilized in the aerospace industry. One such use is to provide flight dynamics models for use in flight simulators. For human-in-the-loop use, such simulators must run in real-time. Owing to the complex physics of rotorcraft flight, to meet this real-time requirement, simplifications to the underlying physics sometimes have to be applied to the model, leading to errors in the model's predictions of the real vehicle's response. This study investigated whether a machine-learning technique could be employed to provide rotorcraft dynamic response predictions. Machine learning was facilitated using a Gaussian process (GP) nonlinear autoregressive model, which predicted the on-axis pitch rate, roll rate, yaw rate, and heave responses of a Bo105 rotorcraft. A variational sparse GP model was then developed to reduce the computational cost of implementing the approach on large datasets. It was found that both of the GP models were able to provide accurate on-axis response predictions, particularly when the model input contained all four control inceptors and one lagged on-axis response term. The predictions made showed improvement compared to a corresponding physics-based model. The reduction of training data to one-third (rotational axes) or one-half (heave axis) resulted in only minor degradation of the sparse GP model predictions.

Download Full-text

Detecting changes in real-time data: a user’s guide to optimal detection

Philosophical Transactions of The Royal Society A Mathematical Physical and Engineering Sciences ◽

10.1098/rsta.2016.0298 ◽

2017 ◽

Vol 375 (2100) ◽

pp. 20160298 ◽

Cited By ~ 3

Author(s):

P. Johnson ◽

J. Moriarty ◽

G. Peskir

Keyword(s):

Power Systems ◽

Real Time ◽

Systems Engineering ◽

Sequential Analysis ◽

Signal To Noise Ratio ◽

Detection Theory ◽

Change Point Detection ◽

Optimal Detection ◽

Time Data ◽

Detection Delay

The real-time detection of changes in a noisily observed signal is an important problem in applied science and engineering. The study of parametric optimal detection theory began in the 1930s, motivated by applications in production and defence. Today this theory, which aims to minimize a given measure of detection delay under accuracy constraints, finds applications in domains including radar, sonar, seismic activity, global positioning, psychological testing, quality control, communications and power systems engineering. This paper reviews developments in optimal detection theory and sequential analysis, including sequential hypothesis testing and change-point detection, in both Bayesian and classical (non-Bayesian) settings. For clarity of exposition, we work in discrete time and provide a brief discussion of the continuous time setting, including recent developments using stochastic calculus. Different measures of detection delay are presented, together with the corresponding optimal solutions. We emphasize the important role of the signal-to-noise ratio and discuss both the underlying assumptions and some typical applications for each formulation. This article is part of the themed issue ‘Energy management: flexibility, risk and optimization’.

Download Full-text

A Study on Developing Cardiac Signals Recording Framework (CARDIF) Using a Reconfigurable Real-Time Embedded Processor

International Journal of Biomedical and Clinical Engineering ◽

10.4018/ijbce.2019070102 ◽

2019 ◽

Vol 8 (2) ◽

pp. 31-44

Author(s):

Uma Arun ◽

Natarajan Sriraam

Keyword(s):

Real Time ◽

Signal To Noise Ratio ◽

Physiological State ◽

Parameters Estimation ◽

Signal Acquisition ◽

Virtual Instrumentation ◽

Time Data ◽

Embedded Processor ◽

Fidelity Measures ◽

The Time Domain

Due to recent developments in technology, there is a significant growth in healthcare monitoring systems. The most widely monitored human physiological parameters is electrocardiogram (ECG) which is useful for inferring the physiological state of humans. Most of the existing multi-channel ECG acquisition systems were not accessible in resource-constrained settings. This research study proposes a cardiac signal recording framework (CARDIF) using a reconfigurable input-output real-time embedded processor by employing a virtual instrumentation platform. The signal acquisition was configured using Lab VIEW virtual instrumentation block sets. A graphical user interface (GUI) was developed for real-time data acquisition and visualization. The time domain heart rate variability (HRV) statistics were calculated using CARDIF, and the same were compared with a clinical grade 12-channel ECG system. The quality of the acquired signals obtained from the proposed and standard systems was measured and compared by calculating signal-to-noise ratio (SNR). The proposed CARDIF was evaluated qualitatively by visual inspection by a clinician and quantitatively by fidelity measures and vital parameters estimation. The results are quite promising and can be extended for clinical validations.

Download Full-text

A Feed Forward Neural Network Based on Model Output Statistics for Short-Term Hurricane Intensity Prediction

Weather and Forecasting ◽

10.1175/waf-d-18-0173.1 ◽

2019 ◽

Vol 34 (4) ◽

pp. 985-997 ◽

Cited By ~ 7

Author(s):

Kirkwood A. Cloud ◽

Brian J. Reich ◽

Christopher M. Rozoff ◽

Stefano Alessandrini ◽

William E. Lewis ◽

...

Keyword(s):

Neural Network ◽

Real Time ◽

Training Data ◽

Lead Times ◽

Time Data ◽

Feed Forward Neural Network ◽

Hurricane Intensity ◽

Feed Forward ◽

Intensity Prediction ◽

Skill Scores

Abstract A feed forward neural network (FFNN) is developed for tropical cyclone (TC) intensity prediction, where intensity is defined as the maximum 1-min average 10-m wind speed. This deep learning model incorporates a real-time operational estimate of the current intensity and predictors derived from Hurricane Weather Research and Forecasting (HWRF; 2017 version) Model forecasts. The FFNN model is developed with the operational constraint of being restricted to 6-h-old HWRF data. Best track intensity data are used for observational verification. The forecast training data are from 2014 to 2016 HWRF reforecast data and cover a wide variety of TCs from both the Atlantic and eastern Pacific Ocean basins. Cross validation shows that the FFNN increasingly outperforms the operational observation-adjusted HWRF (HWFI) in terms of mean absolute error (MAE) at forecast lead times from 3 to 57 h. Out-of-sample testing on real-time data from 2017 shows the HWFI produces lower MAE than the FFNN at lead times of 24 h or less and similar MAEs at later lead times. On the other hand, the 2017 data indicate significant potential for the FFNN in the prediction of rapid intensification (RI), with RI defined here as an intensification of at least 30 kt (1 kt ≈ 0.51 m s−1) in a 24-h period. The FFNN produces 4 times the number of hits in HWFI for RI. While the FFNN has more false alarms than the HWFI, Brier skill scores show that, in the Atlantic, the FFNN has significantly greater skill than the HWFI and probabilistic Statistical Hurricane Intensity Prediction System RI index.

Download Full-text