Raw Measurement Data Research Articles

Neural networks for calculation of complex fields from far-field intensities in mesoscopic photonic systems

Mesoscopic photonics is built on the foundations of theories from mesoscopic physics of electron propagation, although optical techniques have enabled major strides in experimental research on the topic. Theoretical techniques calculate relevant parameters using wave functions or electric fields inside a sample, while experiments mostly measure intensities in the far field. Ideally, the theoretically calculated and experimentally measured parameters must be on equal footing. Here, we employ deep neural networks that calculate near-field intensities and, subsequently, real and complex fields, from simulated far-field measurements. A fully connected deep neural network is built for one-dimensional systems, while a convolutional neural network is developed for two-dimensional systems. The accuracy of these networks is consistently above 95%. We reveal the improvement in estimation of transport parameters by the predicted near-field data from raw measurement data.

Correlation-based damage detection method using convolutional neural network for civil infrastructure

We present a novel damage detection method named CorCNN that utilizes one-dimensional convolutional neural networks to detect damage based on observed changes in correlation between measurements. CNN architecture is used in the method to automatically extract important information from raw measurement data. A CNN model is trained in an unsupervised manner, eliminating the need for data labeling. An assessment of structural responses to a 20 m full-scale bridge in healthy and damaged conditions is conducted to validate the method. For the investigated problem, hyperparameters are optimised to find the optimal combination. To detect the presence of damage, residuals derived from the discrepancies between the actual data and prediction are analyzed. Additionally, CorCNN is compared to other machine learning methods, including linear regression, artificial neural networks, and random forests, using the given dataset. According to the results, the CorCNN method outperforms other machine learning models in detecting damage to the structure.

Practical Approaches for Acoustic Emission Attenuation Modelling to Enable the Process Monitoring of CFRP Machining

Acoustic emission-based monitoring of the milling process holds the potential to detect undesired damages of fibre-reinforced plastic workpieces, such as delamination or matrix cracking. In addition, abrasive tool wear, tool breakage, or coating failures can be detected. As measurements of the acoustic emission are impacted by attenuation, dispersion, and reflection as it propagates from source to sensor, the waveforms, amplitudes, and frequency content of a wave packet differ depending on the propagation length in the workpiece. Since the distance between acoustic emission sources and a stationary sensor attached to the workpiece changes continually in circumferential milling, the extraction of meaningful information from the raw measurement data is challenging and requires appropriate signal processing and frequency-dependent amplification. In this paper, practical and robust approaches, namely experimentally identified transfer functions and frequency gain parameter tables for attenuation modelling, which in reverse enable the reconstruction of frequency spectra emitted at the acoustic emission source, are presented and discussed. From the results, it is concluded that linear signal processing can largely compensate for the influence of attenuation, dispersion, and reflection on the frequency spectra and can therefore enable acoustic emission based process monitoring.

Multi-frequency hyperthermia characterisation via calorimetry and AC magnetometry measurements

Hyperthermia characterisation of magnetic nanoparticles (MNP) is challenging, with highly variable results reported by different laboratories and large uncertainties associated with the measured values. Here, two experimental methods for probing the heating power of MNP for magnetic field hyperthermia (MFH) therapy are compared. The first technique is calorimetry using a bespoke non-adiabatic system capable of measurements at varying frequencies and intensities of magnetic drive-field. Heat losses during calorimetry were compensated for by incorporating a loss-parameter and fitting process during analysis of the raw measurement data to calculate the sample’s specific loss power (SLP). The second technique was AC magnetometry to measure the sample hysteresis when exposed to drive-fields with variable frequencies and amplitudes. The area of the measured hysteresis loop is used to calculate the SLP. Two distinct commercial magnetic nanoparticle systems (Ferucarbotran and RCL-01) were studied using both measurement techniques at varying drive-field frequencies and intensities. Estimates of the intrinsic loss power (ILP) parameter were made based on detailed analysis of the SLP values for multiple frequencies and intensities for each nanoparticle system and experimental approach. For both nanoparticle systems, the calorimetry technique produced higher estimates of the ILP than the AC magnetometry technique. Despite these variations, the results were encouragingly similar, with all values falling within the calculated uncertainty envelopes and relative differences in the results for the two nanoparticles systems being consistent between the measurement techniques (suggesting systematic errors). Furthermore, the experimentally determined values showed strong agreement with literature. The presented study highlights the potential value of carrying out hyperthermia characterisations with differing drive-field types and measurement methods. This approach can reduce inaccuracies by assisting in identifying outliers and aiding the identification of measurements which fall within the Linear-Response Regime for the given nanoparticle. Thus, the presented approach offers additional security when characterising the ILP of MNP.

Development and Assessment of a Movement Disorder Simulator Based on Inertial Data.

The detection analysis of neurodegenerative diseases by means of low-cost sensors and suitable classification algorithms is a key part of the widely spreading telemedicine techniques. The choice of suitable sensors and the tuning of analysis algorithms require a large amount of data, which could be derived from a large experimental measurement campaign involving voluntary patients. This process requires a prior approval phase for the processing and the use of sensitive data in order to respect patient privacy and ethical aspects. To obtain clearance from an ethics committee, it is necessary to submit a protocol describing tests and wait for approval, which can take place after a typical period of six months. An alternative consists of structuring, implementing, validating, and adopting a software simulator at most for the initial stage of the research. To this end, the paper proposes the development, validation, and usage of a software simulator able to generate movement disorders-related data, for both healthy and pathological conditions, based on raw inertial measurement data, and give tri-axial acceleration and angular velocity as output. To present a possible operating scenario of the developed software, this work focuses on a specific case study, i.e., the Parkinson’s disease-related tremor, one of the main disorders of the homonym pathology. The full framework is reported, from raw data availability to pathological data generation, along with a common machine learning method implementation to evaluate data suitability to be distinguished and classified. Due to the development of a flexible and easy-to-use simulator, the paper also analyses and discusses the data quality, described with typical measurement features, as a metric to allow accurate classification under a low-performance sensing device. The simulator’s validation results show a correlation coefficient greater than 0.94 for angular velocity and 0.93 regarding acceleration data. Classification performance on Parkinson’s disease tremor was greater than 98% in the best test conditions.

Exploration of organic superionic glassy conductors by process and materials informatics with lossless graph database

Data-driven material exploration is a ground-breaking research style; however, daily experimental results are difficult to record, analyze, and share. We report a data platform that losslessly describes the relationships of structures, properties, and processes as graphs in electronic laboratory notebooks. As a model project, organic superionic glassy conductors were explored by recording over 500 different experiments. Automated data analysis revealed the essential factors for a remarkable room temperature ionic conductivity of 10−4–10−3 S cm−1 and a Li+ transference number of around 0.8. In contrast to previous materials research, everyone can access all the experimental results, including graphs, raw measurement data, and data processing systems, at a public repository. Direct data sharing will improve scientific communication and accelerate integration of material knowledge.

Robust unsupervised learning of temporal dynamic vehicle-to-vehicle interactions

Robust unsupervised learning of temporal dynamic interactions is an important problem in robotic learning in general and automated unsupervised learning in particular. Temporal dynamic interactions can be described by (multiple) geometric trajectories in a suitable space over which unsupervised learning techniques may be applied to extract useful features from raw and high-dimensional data measurements. Taking a geometric approach to robust unsupervised learning for temporal dynamic interactions, it is necessary to develop suitable metrics and a systematic methodology for comparison and for assessing the stability of an unsupervised learning method concerning its tuning parameters. Such metrics must account for (geometric) constraints in the physical world as well as uncertainties associated with the learned patterns. In this paper, we introduce a model-free metric based on the Procrustes distance for robust learning of dynamic vehicle interactions, and an optimal transport-based distance metric for comparing between distributions of interaction patterns. These distance metrics can serve as an objective for assessing the stability of an interaction learning algorithm. They are also used for comparing the outcomes produced by different algorithms. Moreover, they may also be adopted as an objective function to obtain clusters and representative interaction primitives. These concepts and techniques will be introduced, along with mathematical properties, while their usefulness will be demonstrated in unsupervised learning of vehicle-to-vehicle interactions extracted from the Safety Pilot database, one of the world’s largest databases for connected vehicles. Results reveal that the proposed methods outperform existing techniques under several well-founded notions of clustering efficiencies.

Multitarget Tracking Based on Dynamic Bayesian Network With Reparameterized Approximate Variational Inference

Multitarget tracking (MTT) is an important component of situation-awareness based on the Internet of Things (IoT). Existing algorithms mainly focus on tracking based on conventional measurements, e.g., bearings or ranges. However, measurement parameter estimations are considered in isolation, limiting the accuracy and resolution of MTT, and the related data association is an NP-hard multidimensional assignment problem. In this article, we develop a new one-step MTT algorithm based on a novel dynamic Bayesian network (DBN), i.e., DBNMTT. The new MTT algorithm directly infers target states from the raw measurement data by fusing the array signal model, the signal propagation model, and the motion model. In this new DBNMTT framework, we treat target states and conventional measurements, such as bearings and target energies as hidden random variables. The posterior joint probability optimization problem is translated into the problem of graphical model learning. In this way, we can improve the accuracy and resolution of MTT and convert the NP-hard data association problem to a hidden variable learning problem. For nonconjugate models in the DBNMTT, we develop a novel reparameterized approximation variational inference (ReAVI) approach to solve the learning problem. The ReAVI converts nonconjugate models to conjugate models with new parameters and reuses the mean-field algorithm. The performance of our proposed new MTT method, namely, DBNMTT based on ReAVI (DBNMTT-ReAVI), is analyzed on extensive simulations in challenging scenarios. The simulation results show that the DBNMTT-ReAVI algorithm is superior to conventional measurement-based MTT algorithms in several aspects, including the success probability, convergence, resolution, and accuracy.

Dataset collection from a SubT environment

This article presents a dataset collected from the subterranean (SubT) environment with a current state-of-the-art sensors required for autonomous navigation. The dataset includes sensor measurements collected with RGB, RGB-D, event-based and thermal cameras, 2D and 3D lidars, inertial measurement unit (IMU), and ultra wideband (UWB) positioning systems which are mounted on the mobile robot. The overall sensor setup will be referred further in the article as a data collection platform. The dataset contains synchronized raw data measurements from all the sensors in the robot operating system (ROS) message format and video feeds collected with action and 360 cameras. A detailed description of the sensors embedded into the data collection platform and a data collection process are introduced. The collected dataset is aimed for evaluating navigation, localization and mapping algorithms in SubT environments. This article is accompanied with the public release of all collected datasets from the SubT environment. Link: Dataset

Strain rates for the southern Siberian craton derived from GPS measurements

This paper presents quantitative estimates of the strain rates for the southern part of the Siberian craton derived from GPS data. The obtained results confirmed the theoretical concept of low strain rates (10–9×year–1) in the intra-platform areas. The paper is based on raw measurement data from open sources, previously reported displacement velocities, and measurement data from the Laboratory of Recent Geodynamics of the Institute of the Earth's Crust SB RAS. It is shown that of the 5 mentioned EFT-CORS stations, the BRTK site (Bratsk city) cannot be used to estimate the longterm velocity of the Siberian craton.

Damage Detection and Localization under Variable Environmental Conditions Using Compressed and Reconstructed Bayesian Virtual Sensor Data.

Structural health monitoring (SHM) with a dense sensor network and repeated vibration measurements produces lots of data that have to be stored. If the sensor network is redundant, data compression is possible by storing the signals of selected Bayesian virtual sensors only, from which the omitted signals can be reconstructed with higher accuracy than the actual measurement. The selection of the virtual sensors for storage is done individually for each measurement based on the reconstruction accuracy. Data compression and reconstruction for SHM is the main novelty of this paper. The stored and reconstructed signals are used for damage detection and localization in the time domain using spatial or spatiotemporal correlation. Whitening transformation is applied to the training data to take the environmental or operational influences into account. The first principal component of the residuals is used to localize damage and also to design the extreme value statistics control chart for damage detection. The proposed method was studied with a numerical model of a frame structure with a dense accelerometer or strain sensor network. Only five acceleration or three strain signals out of the total 59 signals were stored. The stored and reconstructed data outperformed the raw measurement data in damage detection and localization.

Radiometric enhancement of full-waveform airborne laser scanner data for volumetric representation in environmental applications

The scientific investigation of forest ecosystems requires precise information on the three-dimensional structure of trees. Full-waveform airborne laser scanner data contain very valuable information on the biophysical structure in forest stands. Beyond 3D point cloud representations obtained from full-waveform decomposition techniques, volumetric representations of full-waveform laser scanner data (i.e. 3D voxel spaces) form a valuable basis for various scientific analysis ranging from biomass determination (or biomass change determination) over habitat analysis to meteorological gas exchange investigations. In order to achieve volumetric forest stand representations of high geometric and radiometric quality, the raw measurement data have to be prepared for the transformation into the voxel structure. The preparation of the full-waveform airborne laser scanner data comprises on the one hand the reconstruction of the differential backscatter cross section. On the other hand, we have to apply a radiometric correction on the reconstructed differential backscatter cross section, since each individual laser pulse echo is significantly affected by attenuation effects caused by partial reflections in higher regions of the crown during the laser pulse propagation through the vegetation. As a result, the structure in the lower parts of the vegetation is underrepresented in the digitized waveform and consequently also in the volumetric reconstruction.In this paper, we present novel methods for the radiometric enhancement of full-waveform airborne laser scanner data for volumetric representations in environmental applications. Our approach allows the numerically stable reconstruction of the effective differential backscatter cross section using appropriate deconvolution and regularization techniques. Moreover, we developed a correction method, which compensates for the loss of signal strength caused by partial reflections on the path of a laser pulse through the canopy. The correction term is derived from the differential backscatter cross section directly via a waveform history analysis. The basic idea of the correction method is a stepwise increase of the signal amplitudes depending on the individual history of each laser pulse. Therefore, the procedure is also referred to as pulse history correction.Our novel methods contribute to an improved access to the information on vegetation structure contained in full-waveform laser scanner data. Furthermore, they allow to overcome limitations of existing approaches, which are mainly based on the extraction of discrete maxima.

Three-dimensional data-driven magnetostatic field computation using real-world measurement data

PurposeThe purpose of this paper is to present the applicability of data-driven solvers to computationally demanding three-dimensional problems and their practical usability when using real-world measurement data.Design/methodology/approachInstead of using a hard-coded phenomenological material model within the solver, the data-driven computing approach reformulates the boundary value problem such that the field solution is directly computed on raw measurement data. The data-driven formulation results in a double minimization problem based on Lagrange multipliers, where the sought solution must conform to Maxwell’s equations while at the same time being as close as possible to the available measurement data. The data-driven solver is applied to a three-dimensional model of a direct current electromagnet.FindingsNumerical results for data sets of increasing cardinality verify that the data-driven solver recovers the conventional solution. Additionally, the practical usability of the solver is shown by using real-world measurement data. This work concludes that the data-driven magnetostatic finite element solver is applicable to computationally demanding three-dimensional problems, as well as in cases where a prescribed material model is not available.Originality/valueAlthough the mathematical derivation of the data-driven problem is well presented in the referenced papers, the application to computationally demanding real-world problems, including real measurement data and its rigorous discussion, is missing. The presented work closes this gap and shows the applicability of data-driven solvers to challenging, real-world test cases.

Geometric Segmentation of 3D Scanned Surfaces for Multi-Sensor Coordinate Metrology

In modern industry, multi-sensor metrology methods are increasingly applied for fast and accurate 3D data acquisition. These method typically start with fast initial digitization by an optical digitizer, the obtained 3D data is analyzed to extract information to provide guidance for precise re-digitization and multi-sensor data fusion. The raw output measurement data from optical digitizer is dense unsorted points with defects. Therefore a new method of analysis has to be developed to process the data and prepare it for metrological verification. This article presents a novel algorithm to manage measured data from optical systems. A robust edge-points recognition method is proposed to segment edge-points from a 3D point cloud. The remaining point cloud is then divided into different patches by applying the Euclidean distance clustering. A simple RANSAC-based method is used to identify the feature of each segmented data patch and derive the parameters. Subsequently, a special region growing algorithm is designed to refine segment the under-segmentation regions. The proposed method is experimentally validated on various industrial components. Comparisons with state-of-the-art methods indicate that the proposed method for feature surface extraction is feasible and capable of achieving favorable performance and facilitating automation of industrial components.

Identification of external load information using distributed optical fiber sensors embedded in an existing road pavement

This paper describes field applications of distributed optical fiber sensors (DOFS) for monitoring the traffic load on an existing road. To investigate the optimal embedment depth for DOFS for pavement monitoring, a series of field tests was conducted by considering three embedment depths (10, 30, and 40 mm) and three moving loads (pedestrian, car and truck). The main objectives were to examine the performance of DOFS embedded in the pavement and investigate its suitability and sensitivity for obtaining information about moving loads at various depths from the road surface. Among the existing backscattering light technology, Rayleigh scattering-based Optical Frequency Domain Reflectometry (OFDR) was chosen in order to achieve a measurement of high-speed, high range, and good repeatability with a very high spatial resolution of 2.56 mm, all of which are necessary for monitoring dynamic responses due to moving loads. In addition, a data-processing method was developed based on a combination of a Hampel identifier and a low-pass filter, which proved to be effective in removing outliers and high-frequency noise from the raw measurement data, respectively. The results showed that it is not only possible to detect a moving load on a road, but that a wide range of information can also be obtained from the strains measured by DOFS, including the type of load, its speed, its weight, the number of axles, the axle spacings (in the case of a vehicle) and the traffic flow. With low maintenance costs, DOFS based on Rayleigh scattering OFDR could be used for collection of accurate and reliable traffic data when embedded in a pavement.

Generative Adversarial Network-Based Scheme for Diagnosing Faults in Cyber-Physical Power Systems.

This paper presents a novel diagnostic framework for distributed power systems that is based on using generative adversarial networks for generating artificial knockoffs in the power grid. The proposed framework makes use of the raw data measurements including voltage, frequency, and phase-angle that are collected from each bus in the cyber-physical power systems. The collected measurements are firstly fed into a feature selection module, where multiple state-of-the-art techniques have been used to extract the most informative features from the initial set of available features. The selected features are inputs to a knockoff generation module, where the generative adversarial networks are employed to generate the corresponding knockoffs of the selected features. The generated knockoffs are then fed into a classification module, in which two different classification models are used for the sake of fault diagnosis. Multiple experiments have been designed to investigate the effect of noise, fault resistance value, and sampling rate on the performance of the proposed framework. The effectiveness of the proposed framework is validated through a comprehensive study on the IEEE 118-bus system.

Design of a Generic Virtual Measurement Workflow for Processing Archived Point Cloud of Trees and Its Implementation of Light Condition Measurements on Stems

Virtual measurement workflow (VMW) was a generic data mining method developed in this study. It was used to extract tree information from archived point clouds under limited conditions by applying virtual measurements in virtual reality. As an example of how to use VMW for a specific topic, the VMW implementation of light condition measurement was further developed. This implementation could measure the temporal and spatial distribution of sunlight on virtual trees (stems). The output was expected as a new type of raw measurement data for tree morphology and phycological studies. At a single tree scale, it facilitated the quantitative interpretation of the growth strategy of branches. By measuring a single tree, it was found that only 4.34% of the stem surface could be illuminated throughout the day (8 h). Meanwhile, 35.87% of the stem surfaces were exposed to sunlight for less than one hour a day. A further mathematical processing of the output, i.e., γ (a ratio between relative area of triangles and relative quantities of triangles in each exposure duration group) improved the sensitivity of identifying differences in lighting conditions. Furthermore, we measured virtual trees of four species from an additional data source using a standardized setting. These include the sessile oak, gemu tree, Masson’s pine, and cherry tree. It was found that the shape of the crown was also significant for the distribution of solar energy on stems. For instance, the gemu tree had a cylindrical tree crown with narrow tree skeleton. A percentage of 10.38 of the surface on the gemu tree was illuminated throughout the day (8 h). The Masson’s pine had similar height and DBH with the gemu tree. However, the elliptical tree crown of the Masson’s pines prevented more lights. The area on the stem that was exposed to sunlight (8 h) dropped from 10.38% to 5.71%. This good differentiation of different crown structures might help this VMW implementation to continue to develop as a tool for identifying the effect of various crown shapes on radiosity for different tree species. The successful development of this VMW implementation had several practical applications for tree studies. Meanwhile, it demonstrated the overall feasibility of VMW and provided a paradigm for further development of other VMW implementations.

Two-step training deep learning framework for computational imaging without physics priors.

Deep learning (DL) is a powerful tool in computational imaging for many applications. A common strategy is to use a preprocessor to reconstruct a preliminary image as the input to a neural network to achieve an optimized image. Usually, the preprocessor incorporates knowledge of the physics priors in the imaging model. One outstanding challenge, however, is errors that arise from imperfections in the assumed model. Model mismatches degrade the quality of the preliminary image and therefore affect the DL predictions. Another main challenge is that many imaging inverse problems are ill-posed and the networks are over-parameterized; DL networks have flexibility to extract features from the data that are not directly related to the imaging model. This can lead to suboptimal training and poorer image reconstruction results. To solve these challenges, a two-step training DL (TST-DL) framework is proposed for computational imaging without physics priors. First, a single fully-connected layer (FCL) is trained to directly learn the inverse model with the raw measurement data as the inputs and the images as the outputs. Then, this pre-trained FCL is fixed and concatenated with an un-trained deep convolutional network with a U-Net architecture for a second-step training to optimize the output image. This approach has the advantage that does not rely on an accurate representation of the imaging physics since the first-step training directly learns the inverse model. Furthermore, the TST-DL approach mitigates network over-parameterization by separately training the FCL and U-Net. We demonstrate this framework using a linear single-pixel camera imaging model. The results are quantitatively compared with those from other frameworks. The TST-DL approach is shown to perform comparable to approaches which incorporate perfect knowledge of the imaging model, to be robust to noise and model ill-posedness, and to be more robust to model mismatch than approaches which incorporate imperfect knowledge of the imaging model. Furthermore, TST-DL yields better results than end-to-end training while suffering from less overfitting. Overall, this TST-DL framework is a flexible approach for image reconstruction without physics priors, applicable to diverse computational imaging systems.

The importance of specifically adsorbed ions for electrokinetic phenomena: Bridging the gap between experiments and MD simulations

Molecular Dynamics (MD) simulations are uniquely suitable for providing molecular-level insights into the Electric Double Layer (EDL) that forms when a charged surface is in contact with an aqueous solution. However, simulations are only as accurate in predicting EDL properties as permitted by the atomic interaction models. Experimental ζ-potential values and surface charges could provide a potentially suitable reference to validate and tune the interaction models, if not for the fact that they themselves are a product of imperfect models used to interpret the raw measurement data. Here, we present an approach to tune an interaction model by comparing Electro-Osmotic Flow (EOF) MD simulations against experimental Streaming Current (SC) measurements while minimizing potential modeling errors arising from both approaches. The point that is least susceptible to interpretation and modeling errors is argued to be at the concentration for which zero flow velocity is observed in EOF simulations and a net zero electric current is measured in SC experiments. At this concentration, the ζ-potential is also zero. We were able to match the experimental concentration at which ζ = 0 in MD simulations for a CaCl2 solution at pH 7.5 in contact with fused silica by tuning the ion-surface Lennard-Jones cross interactions. These interactions were found to greatly affect the ion distribution within the EDL and particularly the formation of inner-sphere surface-complexes, which, in turn, affects the electrokinetic flow. With the ion distribution determined explicitly, a series of properties can be calculated unambiguously, such as the capacitance needed for surface complexation models.

The flow behavior of raw Jojoba oil in comparison with some traditional lube oils

To evaluate the possibility of utilizing the raw Jojoba oils for lubricating purposes in internal combustion engines, the flow behavior of raw Jojoba oil was investigated experimentally. Three customary lube oils of Diesel-, Petrol-, and Motorcycle-engines were employed in comparison with the raw Jojoba oil. Rheometer type of Fann Model of 50SL with coaxial cylinders device was used for this investigation. This investigation reveals that the shear stress of the raw Jojoba oil increases gradually with shear rate in the range of 10–500 s−1 in a linear relationship. At constant shear rate, the shear stresses of the raw Jojoba oil decrease steadily with temperature range of 30−90 C°. The well-known Herschel-Bulkley model can be used to describe sufficiently the flow behavior of the raw Jojoba oil measurement data. The raw Jojoba oil shows flow behavior index equals almost one with slight apparent yield stress. The apparent flow viscosity of the raw Jojoba oil can be analyzed by Arrhenius relationship to show the effect of the applied temperature. At higher operating temperature of 90 °C the raw Jojoba oil exhibits better performance and show higher viscosity values than all the examined lube oils within the range of 10–100 s−1 with viscosity index of 233 which is much higher than the commercial examined oils (range 130–140).