Real Measurement Data Research Articles

Detecting Anomalous Multivariate Time-Series via Hybrid Machine Learning

This article investigates the use of hybrid machine learning (HML) for the detection of anomalous multivariate time-series (MVTS). Focusing on a specific industrial use-case from geotechnical engineering, where hundreds of MVTS need to be analyzed and classified, has permitted extensive testing of the proposed methods with real measurement data. The novel hybrid anomaly detector combines two means for detection, creating redundancy and reducing the risk of missing defective elements in a safety relevant application. The two parts are: 1) anomaly detection based on approximately 50 physics-motivated key performance indicators (KPIs) and 2) an unsupervised variational autoencoder (VAE) with long short-term memory layers. The KPI captures expert knowledge on the properties of the data that infer the quality of produced elements; these are used as a type of auto-labeling. The goal of the extension using machine learning (ML) is to detect anomalies that the experts may not have foreseen. In contrast to anomaly detection in streaming data, where the goal is to locate an anomaly, each MVTS is complete in itself at the time of evaluation and is categorized as anomalous or nonanomalous. The article compares the performance of different VAE architectures [e.g., long short-term memory (LSTM-VAE) and bidirectional LSTM (BiLSTM-VAE)]. The results of using a genetic algorithm to optimize the hyperparameters of the different architectures are also presented. It is shown that modeling the industrial process as an assemblage of subprocesses yields a better discriminating power and permits the identification of interdependencies between the subprocesses. Interestingly, different autoencoder architectures may be optimal for different subprocesses; here two different architectures are combined to achieve superior performance. Extensive results are presented based on a very large set of real-time measurement data.

Enhancing Forward-Looking Image Resolution: Combining Low-Rank and Sparsity Priors

Compressed sensing (CS)-based imaging technology has attracted a lot of interest because it can enhance imaging resolution. Targets of interest for forward-looking imaging radar are typically few in comparison to the entire imaging region. This sparsity allows for the natural application of CS to the reconstruction of high-resolution forward-looking images. However, conventional CS-based imaging methods can only perform well when the signal-to-noise ratio (SNR) is high. Strong noise in radar imaging prevents the CS-based methods from producing excellent imaging results. Inspired by the low-rank property of the received radar target echo and the sparsity of the forward-looking image targets, we present a combined low-rank and sparse prior restricted model for forward-looking imaging with a multichannel array radar to overcome strong noise. To solve the low-rank joint sparse double prior constraint optimization problem, an augmented Lagrange multiplier (ALM) reconstruction method under the framework of the alternating direction multiplier method (ADMM) is proposed. Finally, the results of simulation and real measurement data indicate that our presented method is fairly effective at enhancing the azimuth resolution and robustness of forward-looking radar imaging in comparison to other current methods.

3DInvNet: A Deep Learning-Based 3D Ground-Penetrating Radar Data Inversion

The reconstruction of the 3D permittivity map from ground-penetrating radar (GPR) data is of great importance for mapping subsurface environments and inspecting underground structural integrity. Traditional iterative 3D reconstruction algorithms suffer from strong non-linearity, ill-posedness, and high computational cost. To tackle these issues, a 3D deep learning scheme, called 3DInvNet, is proposed to reconstruct 3D permittivity maps from GPR C-scans. The proposed scheme leverages a prior 3D convolutional neural network with a feature attention mechanism to suppress the noise in the C-scans due to subsurface heterogeneous soil environments. Then a 3D U-shaped encoder-decoder network with multi-scale feature aggregation modules is designed to establish the optimal inverse mapping from the denoised C-scans to 3D permittivity maps. Furthermore, a three-step separate learning strategy is employed to pre-train and fine-tune the networks. The proposed scheme is applied to numerical simulation as well as real measurement data. The quantitative and qualitative results show the networks’ capability, generalizability, and robustness in denoising GPR C-scans and reconstructing 3D permittivity maps of subsurface objects.

Method for improving the accuracy of calculating the tropospheric delay in satellite radio navigation using Kalman filtering

A method is proposed for improving the accuracy of calculating the tropospheric delay in satellite radio navigation using Kalman filtering with a combination of calculation using one of the standard troposphere models. The accuracy of the method was tested on real measurement data in the absence of information about weather conditions. It is concluded that the method can work in real time. The method was implemented and tested in the navigation receiver used by the Farvater company.

Study of Deformation Effect on Land Administration Data in Bali Province

Over time, the earth’s plate moves actively and dynamically, causing crustal deformation. Deformation can occur linearly with a constant movement value every year and also nonlinear due to sudden plate movements that cause tectonic activity. In the land administration scope, the deformation can cause changes in the shape, position, and dimensions of measurement and mapping objects. Currently, there are several potential earthquakes near Bali Island. This research was conducted to investigate the changes that occur in measuring and mapping objects (land administration) in Bali Province. The land administration data employed in this study is the simulated gridding data and the real field measurement data. The deformation data used is based on the deformation model data created by the Geospatial Information Agency (BIG), which accommodates the linear movement model-based block motion model. Based on the simulated velocity, the deformation over 14 years caused horizontal displacement up to 86.2 cm. In addition, sudden movements due to an earthquake and its following postseismic deformation also can affect the changes in the position of land administration data. This change in position has potential problems in mapping, such as gaps and overlaps, without including a proper time-dependent coordinate reference system that accommodates the deformation model.

Deep Learning for Path Loss Prediction at 7 GHz in Urban Environment

In the 6 GHz spectrum sharing band, unlicensed devices are managed by automated frequency coordination (AFC) systems to protect incumbent services from interference. Thus, it is important to select accurate propagation models for interference calculation and analysis. This paper utilizes a model-aided deep learning technique for path loss prediction at 7 GHz, as a representative frequency within the 6 GHz band, in an urban environment. The proposed model is a hybrid model, which leverages both domain expert knowledge from a physics-based general-purposed channel model as well as the learning-based capability from a neural network, for path loss prediction. The model is trained and tested using sufficient-quantity and high-quality real propagation measurement data collected in four locations in an urban environment. Numerical results show that the deep learning model provides a better prediction performance than most empirical models. Furthermore, the feasibility of proposed model generalization to new locations after fine-tuning is examined.

Self-Correcting Algorithm for Estimated Time of Arrival of Emergency Responders on the Highway

Edge computing is one of the key features of the 5G technology-scape that is realizing new and enhanced automotive use cases for improving road safety and emergency response management. Back Situation Awareness (BSA) is such a use case that provides an advance notification to the vehicles of an arriving emergency vehicle (EmV). This paper presents an algorithm for enhancing the accuracy of the advanced Estimated Time of Arrival (ETA) notification of an approaching EmV towards the other vehicles on the highway. The notification is expected to ensure timely reaction by the vehicles to create a clear corridor for the EmV to pass through unhindered, thereby saving critical time to reach the emergency event in a safe manner. The main features of the presented solution are i) the self-correcting algorithm, ii) adaptive and dynamic dissemination areas size allocation, as a response to traffic changes, and iii) the evaluation of the ETA estimation accuracy. We have used the real travel time data measurements collected on the E313 highway (Antwerp, Belgium), to evaluate the performance of the algorithm. The performance is evaluated and compared in terms of accuracy and run-time complexity, using different methods such as Kalman filter, Filter-less method, Moving Average, and Exponential Moving Average filters. It is observed that the Kalman filter provides better accuracy on the ETA estimation, thereby reducing the estimation error by around 14% on average.

Evaluation of Perception Sensor Model Performance in Simulation based on Experimental Findings

The market introduction of automated driving functions poses a significant challenge in terms of safety validation, primarily due to the complexity of real traffic test cases. While virtual testing has emerged as a viable solution, the accurate replication of physical perception sensors in virtual environments remains a formidable task. This study presents an evaluation method for virtual perception sensors, specifically focusing on their performance in automated driving and real-driving behaviour scenarios. Real driving data from a proving ground is collected and integrated into a multi-body simulation software, employing a specialized toolbox for seamless conversion of recorded measurements into simulation-ready data. Virtual sensor models for LIDAR, Radar, and Camera, utilizing various machine learning approaches, are implemented within the simulation alongside commercial sensor models. The output of these models is compared to real measurement data using statistical metrics, including the Chebyshev Distance, Pearson Correlation Coefficient, and Cross-Correlation Coefficient. The evaluation highlights the accuracy and performance of the machine learning-based models and the importance of employing multiple metrics that consider both correlation and offset between simulated and measured data.

Reduced-Complexity Subarray-Level Sparse Recovery STAP for Multichannel Airborne Radar WGMTI Application

For multichannel airborne radars, wide-area ground-moving target indication (WGMTI) processing can quickly obtain the dynamic distribution of moving targets in a wide area, which holds considerable significance in many fields. Nevertheless, the WGMTI mode suffers from the interference of powerful ground clutter, which frequently submerges slow-moving targets and causes many false alarms in subsequent moving target detection. Space–time adaptive processing (STAP) can successfully suppress clutter, but its performance depends critically on the available training samples. Consequently, an effective STAP method characterized by fast processing and a small sample size for WGMTI application in multichannel airborne radars must be developed. In this paper, a subarray-level sparse recovery STAP (SR-STAP) processing framework is proposed for multichannel airborne radars. First, the characteristics of the subarray-level received clutter are discussed in detail. Second, on the basis of this analysis, we further designed a joint space–time dictionary and developed a separable tensor-based sparse Bayesian learning (STSBL) method. In this method, two-stage decomposition is proposed to ensure that large-scale data can be degraded into small-scale data in processing, which significantly improves computation efficiency. Finally, the effectiveness of the proposed STSBL-STAP method in WGMTI processing was verified using real measurement data obtained from a developed dual-channel Ku-band airborne radar.

Monitoring Vehicle Pollution and Fuel Consumption Based on AI Camera System and Gas Emission Estimator Model.

Road traffic is responsible for the majority of air pollutant emissions in the cities, often presenting high concentrations that exceed the limits set by the EU. This poses a serious threat to human health. In this sense, modelling methods have been developed to estimate emission factors in the transport sector. Countries consider emission inventories to be important for assessing emission levels in order to identify air quality and to further contribute in this field to reduce hazardous emissions that affect human health and the environment. The main goal of this work is to design and implement an artificial intelligence-based (AI) system to estimate pollution and consumption of real-world traffic roads. The system is a pipeline structure that is comprised of three fundamental blocks: classification and localisation, screen coordinates to world coordinates transform and emission estimation. The authors propose a novel system that combines existing technologies, such as convolutional neural networks and emission models, to enable a camera to be an emission detector. Compared with other real-world emission measurement methods (LIDAR, speed and acceleration sensors, weather sensors and cameras), our system integrates all measurements into a single sensor: the camera combined with a processing unit. The system was tested on a ground truth dataset. The speed estimation obtained from our AI algorithm is compared with real data measurements resulting in a 5.59% average error. Then these estimations are fed to a model to understand how the errors propagate. This yielded an average error of 12.67% for emitted particle matter, 19.57% for emitted gases and 5.48% for consumed fuel and energy.

BEM-Based Magnetic Field Reconstruction by Ensemble Kálmán Filtering

Abstract Magnetic fields generated by normal or superconducting electromagnets are used to guide and focus particle beams in storage rings, synchrotron light sources, mass spectrometers, and beamlines for radiotherapy. The accurate determination of the magnetic field by measurement is critical for the prediction of the particle beam trajectory and hence the design of the accelerator complex. In this context, state-of-the-art numerical field computation makes use of boundary-element methods (BEM) to express the magnetic field. This enables the accurate computation of higher-order partial derivatives and local expansions of magnetic potentials used in efficient numerical codes for particle tracking. In this paper, we present an approach to infer the boundary data of an indirect BEM formulation from magnetic field measurements by ensemble Kálmán filtering. In this way, measurement uncertainties can be propagated to the boundary data, magnetic field and potentials, and to the beam related quantities derived from particle tracking. We provide results obtained from real measurement data of a curved dipole magnet using a Hall probe mapper system.

Cell Fault Identification and Localization Procedure for Lithium-Ion Battery System of Electric Vehicles Based on Real Measurement Data

Vehicle safety risk can be decreased by diagnosing the lithium-ion battery system of electric road vehicles. Real-time cell diagnostics can avoid unexpected occurrences. However, lithium-ion batteries in electric vehicles can significantly differ in design, capacity, and chemical composition. In addition, the battery monitoring systems of the various vehicles are also diverse, so communication across the board is not available or can only be achieved with significant difficulty. Hence, unique type-dependent data queries and filtering are necessary in most cases. In this paper, a Volkswagen e-Golf electric vehicle is investigated; communication with the vehicle was implemented via an onboard diagnostic port (so-called OBD), and the data stream was recorded. The goal of the research is principally to filter out, identify, and localize defective/weak battery cells. Numerous test cycles (constant and dynamic measurements) were carried out to identify cell abnormalities (so-called deviations). A query and data filtering process was designed to detect defective battery cells. The fault detection procedure is based on several cell voltage interruptions at various loading levels. The methodology demonstrated in this article uses a fault diagnosis technique based on voltage abnormalities. In addition, it employs a hybrid algorithm that executes calculations on measurement and recorded data. In the evaluation, a status line comprising three different categories was obtained by parametrizing and prioritizing (weighting) the individual measured values. It allows the cells to be divided into the categories green (adequate region), yellow (to be monitored), and red (possible error). In addition, several querying strategies were developed accordingly to clarify and validate the measurement results. The several strategies were examined individually and analyzed for their strengths and weaknesses. Based on the results, a data collection, processing, and evaluation strategy for an electric vehicle battery system have been developed. The advantage of the developed algorithm is that the method can be adapted to any electric or hybrid vehicle battery.

Benchmarking of Gaussian Process Regression with Multiple Random Fields for Spatial Variability Estimation

Benchmarking is very valuable for evaluating and comparing methodologies. Here, Gaussian process regression using multiple Gaussian random fields (GPR-MR) is applied to benchmarking data for spatial variability problems. The benchmarking data used were from the literature and included four types of virtual ground models (VG1 to VG4) and one real ground measurement data set. The spatial variability of geological properties is often divided into a trend component and a random component. In GPR-MR, the trend component is expressed by a random field with a large scale of fluctuation (SOF), leading to a smooth (slow) variability, whereas the random component is expressed by one with a small SOF, leading to a rapidly changing variability. The SOF and the standard deviation of random fields were estimated using the maximum likelihood method based on the measured data provided in the benchmarking data. GPR-MR was used to estimate the spatial variabilities of all cases, and its performance was evaluated. For the real ground measured data, model selection was also performed with respect to the autocorrelation function of the random component in terms of information criteria, whereas the Markovian autocorrelation function was used for the virtual ground data without the model selection. Based on the results, the Whittle-Matérn (WM) model was selected for the random component. GPR-MR was used to estimate the spatial variability, and its performance with the WM model was evaluated.

Dynamic Response of Power Systems With Real GICs: Impact on Generator Excitation Control

Geomagneticallyinduced current (GIC) studies usually focus on transformer response to dc but seldom analyze generator excitation systems. A GIC is driven by an electric field which is induced by a very low-frequency geomagnetic field. It is widely acceptable to model GICs with dc, mostly suitable for static voltage stability analysis, but this does not adequately characterize a power system's dynamic response. Considering this, we studied a multi-machine power system on an OPAL-RT digital real-time simulator (DRTS)-MATLAB/Simulink system. We used real GIC measurement data from high and mid-latitude regions. Wavelet analysis related the frequency components of the GIC to the power system's response. A dynamic generator model with excitation voltage control modelled the source. Our main finding is that excitation system control contributes to reducing the voltage drop caused by increased var demand in saturated transformers with GICs. Measured GICs revealed non-negligible short-term dynamics affecting bus voltages that are not captured with dc-modelling assumptions, yet they are important for voltage stability analysis. This study implies that excitation control creates a buffer for high GICs to flow without violating the grid codes for minimum operating voltages and this paper demonstrates the optimization of this novel approach.

ASSESSMENT OF STRING INVERTERS FOR LARGE-SCALE SOLAR POWER PLANT IN MALAYSIA’S

Tremendous promotion and growth of Large-Scale Solar Photovoltaic (LSSPV) Power Plant in Malaysia increase the reliability and potentiality of the system in having efficient monitoring, operation and maintenance for the plants. Through the Ministry of Energy, Science and Technology, the government aims to increase the renewable energy in its generation mix in 2025 by 20% and intends to reduce its greenhouse gas (GHG) emission intensity per unit of gross domestic product (GDP) to 45% by 2030. This was followed by the Energy Commission (EC) announcement of 500 MW LSSPV power plants in Malaysia, slated for commercial operation in 2021. Thus, proper monitoring and respectable power generation are crucial to optimising LSSPV for grid interconnection. This paper presents the assessment of string inverters configuration and analysis models to recommend an optimised configuration for maximum harvested power generation in the LSSPV power plant in Malaysia. An analysis of string invert configuration is carried out based on real data measurements of electric generation of a 50 MW LSSPV power plant at the Kuala Ketil site. The test verified that the configuration string inverters offer a reliable outcome for the grid code test (GCT). Besides, the string inverters provide a potential approach to maximising power generation at different climate conditions, thus, reducing the workforce for the operation and maintenance activities. The string system endeavours a feasibility implementation on the suitability of effective configurations for LSSPV in Malaysia without often shutting down the plant generations due to any defect or damage of PV modules.

Influence of Hydraulic Model Complexity on Results of Water Age and Quality Simulation in Municipal Water Supply Systems

The age of water in the municipal water supply system is one of the main factors influencing water quality. To create a good quality hydraulic model, one must achieve a high level of calibration accuracy with real life measurement data. Before we start building our model, we must decide on the model’s level of detail, that is, its complexity. We must know if skeletonization of the network graph and different hydraulic timesteps have an influence on simulation results. This study strives to prove that this decision can lead to unforeseen problems during the calibration process, thus making it impossible to achieve the required calibration precision. In order to prove this, two different model variants were created with different levels of graph detail, and simulation data results were used to determine which model variant is best suited to achieve the highest fidelity simulation results. Following this, the chosen model was run with different hydraulic timestep settings, which made it possible to showcase the large influence this setting has on achieved results.

Development of High-Fidelity Automotive LiDAR Sensor Model with Standardized Interfaces.

This work introduces a process to develop a tool-independent, high-fidelity, ray tracing-based light detection and ranging (LiDAR) model. This virtual LiDAR sensor includes accurate modeling of the scan pattern and a complete signal processing toolchain of a LiDAR sensor. It is developed as a functional mock-up unit (FMU) by using the standardized open simulation interface (OSI) 3.0.2, and functional mock-up interface (FMI) 2.0. Subsequently, it was integrated into two commercial software virtual environment frameworks to demonstrate its exchangeability. Furthermore, the accuracy of the LiDAR sensor model is validated by comparing the simulation and real measurement data on the time domain and on the point cloud level. The validation results show that the mean absolute percentage error of simulated and measured time domain signal amplitude is . In addition, the of the number of points and mean intensity values received from the virtual and real targets are and , respectively. To the author’s knowledge, these are the smallest errors reported for the number of received points and mean intensity values up until now. Moreover, the distance error is below the range accuracy of the actual LiDAR sensor, which is 2 cm for this use case. In addition, the proving ground measurement results are compared with the state-of-the-art LiDAR model provided by commercial software and the proposed LiDAR model to measure the presented model fidelity. The results show that the complete signal processing steps and imperfections of real LiDAR sensors need to be considered in the virtual LiDAR to obtain simulation results close to the actual sensor. Such considerable imperfections are optical losses, inherent detector effects, effects generated by the electrical amplification, and noise produced by the sunlight.

Integrated Tracking and ISAR Imaging Using an Integrated Kalman Filter With Wideband Radar

Traditionally, inverse synthetic aperture radar (ISAR) imaging and target tracking are two separate and independent processes. ISAR imaging is usually achieved after data collecting and tracking for a long time, resulting in the problems of low imaging efficiency, information loss, and probing resource allocation difficulty. In this article, we proposed an integration method of tracking and imaging with wideband radar by constructing the relation among signal, data, and image domains, which can realize accurate tracking and high data-rate imaging at the same time. In this integrated imaging and tracking method, a complex-valued reference high-resolution range profile (HRRP) is generated and predicted by exploiting the relation among range-Doppler image, echo in time domain, and target’s motion state. Then, a Bayes estimation model is built to extract range phase error between the predicted HRRP and the real-measured HRRP, and the Kalman filter can update the motion state. In addition, in order to increase the efficiency, the decentralized Kalman filter is also introduced to realize the subaperture parallel processing and data fusion in this article. Finally, the sequential motion compensation is carried out to realize accurate motion compensation and sequential imaging. The simulated and real-measured data confirm the proposed algorithm’s robust imaging and tracking performance. Experimental results show that the proposed method has excellent efficiency and can simultaneously realize sequential imaging and accurate tracking.

A Multi-dimensional Real World Spectrum Occupancy Data Measurement and Analysis for Spectrum Inference in Cognitive Radio Network

Spectrum Inference in contrast to Spectrum Sensing is an active technique for dynamically inferring radio spectrum state in Cognitive Radio Networks. Efficient spectrum inference demands real world multi-dimensional spectral data with distinct features. Spectrum bands exhibit varying noise floors; an effective band wise noise thresholding guarantees an accurate occupancy data. In this work, we have done an extensive real world spectrum occupancy data measurement in frequency range 0.7 GHz to 3 GHz for tele density wise varying locations at Pune, Solapur and Kalaburagi with time diversity ranging from 2 to 7 days. We have applied maximum noise (Max Noise), m-dB and probability of false alarm (PFA) noise thresholding for spectrum occupancy calculations in all bands and across all locations. Overall occupancy across these locations is 37.89 %, 18.90 % and 13.69 % respectively. We have studied signal to noise ratio (SNR), channel vacancy length durations (CVLD) and service congestion rates (SCR) as characteristic features of measured multi-dimensional spectrum data. The results reveal strong time, spectral and spatial correlations of these features across all locations. These features can be used for a multi-dimensional spectrum inference in cognitive radio based on machine learning.

Deep learning-based rapid damage assessment of RC columns under blast loading

Reinforced concrete (RC) columns may be exposed to blast loading. Assessment and prediction of the damage condition and failure mode of RC columns are necessary for ensuring the structural safety and taking protective measures. However, rapid damage assessment of RC columns to blasts is challenging due to the lack of real measurement data and time-consuming numerical simulation. In this study, two deep learning models are established and trained in a supervised end-to-end manner to evaluate the damage degree and predict the failure mode of RC columns. To solve the shortage of real data, over 3000 numerical simulation datasets are generated and integrated with over 200 measurement data from the literature as the training data. First, a multi-hidden-layer neural network model is constructed using PyTorch to assess the damage degree of the RC columns under blast loading. Eight RC column structural parameters and two blast load parameters are used as inputs, and the damage index of the RC column is the output. Second, the Keras framework is used to build a multi-layer long-short-term memory neural network model with the same eight parameters as the inputs and the failure mode of the RC columns as the output to predict the failure mode of the RC columns. The established models are also applied to the explosion risk assessment of the RC column in a frame structure. The zoning maps of the damage degree and the failure mode are established. The results show that the developed deep learning models can accurately and rapidly assess the damage degree and predict the failure mode of the RC column under blast loading, which provides guidance for the anti-explosion protection design of RC columns.