Search

Solid oxide fuel cells (SOFC) could facilitate the green energy transition as they can produce high-temperature heat and electricity while emitting only water when supplied with hydrogen. Additionally, when operated with natural gas, these systems demonstrate higher thermoelectric efficiency compared to traditional microturbines or alternative engines. Within this context, although digitalisation has facilitated the acquisition of extensive data for precise modelling and optimal management of fuel cells, there remains a significant gap in developing digital twins that effectively achieve these objectives in real-world applications. Existing research predominantly focuses on the use of machine learning algorithms to predict the degradation of fuel cell components and to optimally design and theoretically operate these systems. In light of this, the presented study focuses on developing digital twin-oriented models that predict the efficiency of a commercial gas-fed solid oxide fuel cell under various operational conditions. This study uses data gathered from an experimental setup, which was employed to train various machine learning models, including artificial neural networks, random forests, and gradient boosting regressors. Preliminary findings demonstrate that the random forest model excels, achieving an R2 score exceeding 0.98 and a mean squared error of 0.14 in estimating electric efficiency. These outcomes could validate the potential of machine learning algorithms to support fuel cell integration into energy management systems capable of improving efficiency, pushing the transition towards sustainable energy solutions. [ABSTRACT FROM AUTHOR]

In recent years, the detection performance of SAR-GMTI (synthetic aperture radar-ground moving target indication) algorithm based on deep learning has always been limited by insufficient measured data due to the heavy operation complexity and high cost of real SAR systems. To solve this problem, this paper proposes an overall DT-based implementation framework for SAR ground moving target intelligent detection tasks. In particular, by virtue of a SAR imaging algorithm, a high-fidelity twin replica of SAR moving targets is established in digital space through parameter traversal based on the prior target characteristics of the obtained measured datasets. Then, the constructed SAR twin datasets is fed into the neural network model to train an intelligent detector by fully learning features of the moving targets and preset the SAR scene in the twin space, which can realize the robust detection of ground moving targets in related practical scenarios with no need for multiple and complex field experiments. Moreover, the effectiveness of the proposed framework is verified on the MiniSAR measured system, and a comparison with traditional CFAR detection method is given simultaneously. [ABSTRACT FROM AUTHOR]

The paper presents a wireless system integrated with a machine learning (ML) model for structural health monitoring (SHM) of carbon fiber reinforced polymer (CFRP) structures, primarily targeting aerospace applications. The system collects data via carbon nanotube (CNT) piezoresistive sensors embedded within CFRP coupons, wirelessly transmitting these data to a central server for processing. A deep neural network (DNN) model predicts mechanical properties and can be extended to forecast structural failures, facilitating proactive maintenance and enhancing safety. The modular design supports scalability and can be embedded within digital twin frameworks, offering significant benefits to aircraft operators and manufacturers. The system utilizes an ML model with a mean absolute error (MAE) of 0.14 on test data for forecasting mechanical properties. Data transmission latency throughout the entire system is less than one second in a LAN setup, highlighting its potential for realtime monitoring applications in aerospace and other industries. However, while the system shows promise, challenges such as sensor reliability under extreme environmental conditions and the need for advanced ML models to handle diverse data streams have been identified as areas for future research. [ABSTRACT FROM AUTHOR]

Emerging from the convergence of digital twin technology and the metaverse, consumer health (MCH) is witnessing a transformative shift. The amalgamation of bioinformatics with healthcare Big Data has ushered in a new era of disease prediction models that harness comprehensive medical data, enabling the anticipation of illnesses even before the onset of symptoms. In this model, deep neural networks stand out because they improve accuracy remarkably by increasing network depth and making weight changes using gradient descent. Nonetheless, traditional methods face their own set of challenges, including the issues of gradient instability and slow training. In this case, the Broad Learning System (BLS) stands out as a good alternative. It gets around the problems with gradient descent and lets you quickly rebuild a model through incremental learning. One problem with BLS is that it has trouble extracting complex features from complex medical data. This makes it less useful in a wide range of healthcare situations. In response to these challenges, we introduce DAE-BLS, a novel hybrid model that marries Denoising AutoEncoder (DAE) noise reduction with the efficiency of BLS. This hybrid approach excels in robust feature extraction, particularly within the intricate and multifaceted world of medical data. Validation using diverse datasets yields impressive results, with accuracies reaching as high as 98.50%. DAE-BLS's ability to rapidly adapt through incremental learning holds great promise for accurate and agile disease prediction, especially within the complex and dynamic healthcare scenarios of today. [ABSTRACT FROM AUTHOR]

Exact observing and forecasting tool conditions fundamentally affect cutting execution, bringing further developed workpiece machining accuracy and lower machining costs. Because of the unpredictability and time-differing nature of the cutting system, existing methodologies cannot achieve ideal oversight progressively. A technique dependent on Digital Twins (DT) is proposed to accomplish extraordinary accuracy in checking and anticipating tool conditions. This technique builds up a balanced virtual instrument framework that matches entirely with the physical system. Collecting data from the physical system (Milling Machine) is initialized, and sensory data collection is carried out. The National Instruments data acquisition system captures vibration data through a uni-axial accelerometer, and a USB-based microphone sensor acquires the sound signals. The data are trained with different Machine Learning (ML) classification-based algorithms. The prediction accuracy is calculated with the help of a confusion matrix with the highest accuracy of 91% through a Probabilistic Neural Network (PNN). This result has been mapped by extracting the statistical features of the vibrational data. Testing has been performed with the trained model to validate the model's accuracy. Later, the modeling of the DT is initiated using MATLAB-Simulink. This model has been created under the data-driven approach. The physical–virtual balance of the DT model is acknowledged utilizing the advances, taking into consideration the detailed planning of the constant state of the tool's condition. The tool condition monitoring system through the DT model is deployed through the machine learning technique. The DT model can predict the different tool conditions based on sensory data. [ABSTRACT FROM AUTHOR]

Irregular ocean waves contribute to random and non-linear hydrodynamic forces on marine and offshore platforms, in particular those made of slender bodies such as jack-ups. Keeping track of the structural fatigue, integrity, and reliability renders benefits to these platforms such as enabling predictive maintenance and risk-based decision support. While computational techniques are useful in providing insights to deepen our understanding of the physics involved, these are resource-intensive and not suited for predictive capabilities. Here we introduce an effective deep learning method based on Long Short Time Memory (LSTM) to predict unsteady hydrodynamic forces on jack-up legs, which are mainly governed by non-linear drag forces, as well as dynamic responses of the jack-up in terms of reaction forces and hull motion. The numerical OF-ABAQUS solver (one-way fluid-structure coupling) designed for jack-up by Li et al. (2022, 2023) is used to generate a database for the deep neural network, which takes in wave elevation time series as the input. A leaky non-linear rectification is used to trace the relationship between wave elevation (as the input) and hydrodynamic forces as well as structural responses (as the outputs). The LSTM model is trained using a stochastic gradient descent method, and the results are compared with the full-order numerical simulations. Once the hyper-parameters of the LSTM algorithm are identified, further tuning for each sea-state is not required. Instead, the same LSTM algorithm can be used for different sea-states. The predictions made with our LSTM model are real-time, 96,000 times faster than the full-order simulations, within a 6% error threshold, and use a negligible amount of central processing unit (CPU) power. This powerful tool can be used to obtain real-time predictions using computers on offshore platforms without high-performance computers inland. Further, the proposed LSTM-based method can be applied to create digital twins for jack-up platforms. [ABSTRACT FROM AUTHOR]

Health monitoring of structures using ultrasonic guided waves is an evolving technology with potential applications in monitoring pipelines, civil bridges, and aircraft components. However, the sensitivity of guided waves to external parameters affects the reliability of monitoring systems based on them. These influencing factors and experimental related factors cannot be perfectly modeled, which give rise to the discrepancy between numerical simulations and experimental measurements. Therefore, it is important to address this inevitable discrepancy and generate close-to-experiment simulations. In this work, we present a deep learning-based Digital Twin framework containing multi-fidelity modeling to reduce the discrepancy between measurements and simulations and a deep generative model to generate close-to-experiment guided wave responses by harnessing the vital characteristics of the two sources. These realistic simulations (close to experiment) can then be used in assessing the reliability of health monitoring system by generating probability of detection curves. Furthermore, they can also be used for augmenting the training data for a machine learning algorithm. We use a measurement dataset corresponding to crack propagation and simulations to validate the proposed framework. The results show that the discrepancy is indeed reduced to a great extent, furthermore, we also show that this framework enables the computation of probability of detection from close-to-experiment data as a direct consequence of rapid generation of realistic simulations. • Structural health monitoring based on ultrasonic guided waves. • Discrepancy reduction between simulations and experiments. • A novel unifed deep neural network model blending simulations and experiments. • Enhanced ultrasonic guided waves signals generation. • Discrepancy quantification through phase and amplitude based misfit metrics. • Metamodel based computation of probability of detection. [ABSTRACT FROM AUTHOR]

Modeling the photovoltaic (PV) energy output with high accuracy is essential for predicting and analyzing the performance of a PV system. In the particular cases of building-integrated and building-attached photovoltaic systems (BIPV and BAPV, respectively) the time-varying partial shading conditions are a relevant added difficulty for modeling the PV power conversion. The availability of laser imaging detection and ranging (LIDAR) data to create very-high-resolution elevation digital models can be effectively used for computing the shading at high resolution. In this work, an artificial neural network (ANN) has been used to model the power generation of different BIPV arrays on a 5 min basis using the meteorological and solar irradiance on-site conditions, as well as the shading patterns estimated from a digital surface model as inputs. The ANN model has been validated using three years of 5-min-basis monitored data showing very high accuracy (6–16% of relative error depending on the façade). The proposed methodology combines the shading computation from a digital surface model with powerful machine learning algorithms for modeling vertical PV arrays under partial shading conditions. The results presented here prove also the capability of the machine learning techniques towards the creation of a digital twin for the specific case of BIPV systems that complements the conventional monitoring strategies and can be used in the diagnosis of performance anomalies. [ABSTRACT FROM AUTHOR]

Real-time weighing plays a crucial role in various applications in the road infrastructure field, such as pavement design, pavement monitoring systems, and the development of digital twins. Innovative sensing technologies offer solutions to the limitations of traditional sensors, enhancing durability and cost-effectiveness. Current research on self-sensing technologies primarily focuses on material manufacturing, electrode embedding, and testing of piezoresistive properties. However, there is limited investigation of innovative methods for processing and analyzing electrical signals from self-sensing materials for load classification. The novelty of the present paper lies in the development of a machine learning-based load classifier using electrical data from laboratory self-sensing asphalt mixtures with carbon fibers (CFs). Specimens underwent repeated cyclic dynamic loads, and electrical signals were recorded and processed using a customized digital processing algorithm. Artificial Neural Network (ANN) and decision tree algorithms were employed to develop load classification models. Results indicate that the addition of CFs strengthens the asphalt mixture, converting it into a piezoresistive material with load-sensing capabilities. Conclusions show that while the machine learning algorithms performed satisfactorily, further research is required for optimization purposes. [Display omitted] • Carbon fibres transform asphalt mixtures into a self-sensing material. • Electromechanical laboratory tests were conducted. • Discrete Wavelet Transform (DWT) was employed for filtering data. • Different machine-learning algorithms were developed for load classification. [ABSTRACT FROM AUTHOR]

Digital twin ushers in growth as one cutting-edge technology that enables high-precision, informative and real-time interaction of advanced composites, influencing the paradigm of composite analysis and design in aerospace, automotive, intelligent electronics and other fields. The online and fine evolution of digital twins at scale face challenges due to creating scalable predictions, updates, and controls with coupled information in workflows. In this work, an interactive digital twin methodology for complex composite structures (CCSDT) based on machine learning is introduced. To demonstrate the framework's feasibility and applicability, a hybrid architecture comprising hierarchical deep neural networks (H-DNNs), statistical inference and cross-scale physical constraints is proposed considering the fabric rubber composites (FRC) in aerospace field. The framework predicts 3D displacement and stress fields directly from sensing features and can incorporate dynamic updating and evaluation of computational models into data assimilation. Real-time prediction and uncertainty quantification are verified through synthetic and experimental data, and the capability of CCSDT in perception and decision is demonstrated by a case of cruise state monitoring. The comparison between the measured strain field and the post-processing predicted strain field shows the extensibility of the direct prediction results. These results provide guidance for the development of composite digital twins, stimulating the potential for cost-effective and efficient digital twin services. [Display omitted] [ABSTRACT FROM AUTHOR]

Monitoring Aboveground Net Primary Production (ANPP) is critical to assess not only the current ecosystem status but also its long-term dynamics. In rangelands, the seasonal dynamics of ANPP determines forage availability, stock density, and livestock productivity. To develop a hybrid model to be used as a prediction engine for ANPP in the native grasslands of Uruguay. The model combines a parametric component based on the seasonal dynamics of ANPP, and an artificial neural network (ANN) component used to model the remaining non-linearities, which are mainly related to precipitation and temperature variability. The output of hybrid model is proposed as the "virtual entity" of a digital twin support decision system where the "physical entity" is characterized by a collection of bi-weekly (fortnight) ANPP estimates. Fortnight ANPP data were calculated from MODIS EVI for the 2001–2020 period. A sigmoidal functional response, having three parameters with an explicit biological interpretation, was fitted to the accumulated ANPP as a function of time. Forecasts were generated by extrapolating the sigmoidal functional response fit up to four fortnights ahead. From these fits, we obtained the fortnight ANPP values by differentiating the accumulated fortnight ANPP. Predictions (up to four fortnights) were generated for each fortnight and year. The residuals from these fits were modeled using a multilayer perceptron trained by backpropagation using climate variables as independent variables. The sigmoidal functional response model fit was highly significant for the accumulated ANPP profile. This model also had a high explanatory power for the accumulated ANPP curve. The median of the percentage absolute residuals for forecasts made 1 to 4 fortnights ahead ranged from 17% to 18%. The ANN significantly reduced this unexplained variability in ANPP, showing a median reduction in residuals of 35%, 31%, 30%, and 30% for 1 to 4 fortnights ahead forecasts, respectively, when compared to predictions from the sigmoidal functional response fit. By integrating both parametric and machine learning techniques, the hybrid model developed can make accurate predictions in a way that is both efficient and dependable. The hybrid model not only represents an advantage in terms of predictive power, but it also allows for a deeper understanding of the basic ecological processes involved in forage production. [Display omitted] • We developed a hybrid model (HM) and prediction engine for Aboveground Net Primary Production (ANPP). • Integrating parametric and machine learning techniques, this HM makes predictions that outperform traditional models. • The hybrid approach also helps in the understanding of basic ecological processes involved in forage production. • Applicability includes planning and optimizing grazing management of native grasslands in commercial farms. • We propose the use of the output of the HM as the "virtual entity" of a digital twin decision system to support grassland management. [ABSTRACT FROM AUTHOR]

Light‐emitting diodes (LEDs) are among the key innovations that have revolutionized the lighting industry, due to their versatility in applications, higher reliability, longer lifetime, and higher efficiency compared with other light sources. The demand for increased lifetime and higher reliability has attracted a significant number of research studies on the prognostics and lifetime estimation of LEDs, ranging from the traditional failure data analysis to the latest degradation modeling and machine learning based approaches over the past couple of years. However, there is a lack of reviews that systematically address the currently evolving machine learning algorithms and methods for fault detection, diagnostics, and lifetime prediction of LEDs. To address those deficiencies, a review on the diagnostic and prognostic methods and algorithms based on machine learning that helps to improve system performance, reliability, and lifetime assessment of LEDs is provided. The fundamental principles, pros and cons of methods including artificial neural networks, principal component analysis, hidden Markov models, support vector machines, and Bayesian networks are presented. Finally, discussion on the prospects of the machine learning implementation from LED packages, components to system level reliability analysis, potential challenges and opportunities, and the future digital twin technology for LEDs lifetime analysis is provided. [ABSTRACT FROM AUTHOR]

• A data-driven method is proposed to evaluate the ultimate response of masonry arches. • Machine learning metamodels, in the form of artificial neural networks, are adopted. • Numerical datasets using Matlab, Python and FEM software are generated. • A fast and accurate prediction of the collapse mode and ultimate load is achieved. • Source Matlab, Python files and generated datasets accompany the article. A data-driven methodology is proposed, for the investigation of the ultimate response of masonry arches. Aiming to evaluate their structural response in a computationally efficient framework, machine learning metamodels, in the form of artificial neural networks, are adopted. Datasets are numerically built, integrating Matlab, Python and commercial finite element software. Heyman's assumptions are adopted within non-linear finite element analysis, incorporating contact-friction laws between adjacent stones, to capture failure in the arch. The artificial neural networks are trained, validated, and tested using the least square minimization technique. It is shown that the proposed scheme can be used to provide a fast and accurate prediction of the deformed geometry, the collapse mechanism and the ultimate load. Cases studies demonstrate the efficiency of the method in random, new arch geometries. Relevant Matlab/Python scripts and datasets are provided. The method can be extended towards structural health monitoring and the concept of digital twin. [ABSTRACT FROM AUTHOR]

• This paper presents a feed-forward neural network for predicting hydroelastic response of VLFS. • The Levenberg-Marquardt optimisation techniques is shown to provide higher prediction accuracy. • The ANN is trained using dataset generated by finite element-boundary element method.The trained machine is used to predict the hydroelastic response under regular and irregular sea. • ANN is useful for applications such as structural health monitoring and digital twin. The response of a very large floating structure (VLFS) must take into consideration the elastic deformation of the structure (commonly termed hydroelastic response) under wave action. Conventionally, the hydroelastic response could be computed by using the coupled finite element-boundary element (FE-BE) method, where the mat-like structure is modelled using plate theory and the water modelled using the potential theory. The FE-BE method requires the structure to be discretised into finer elements and the wetted surface boundary to be represented by smaller panels to accurately capture the hydroelastic response of the structure. Thus, the coupled FE-BE method could be computationally expensive when the structure gets larger or when subjected to waves of smaller wavelengths. To accelerate the computational time in predicting the hydroelastic response of the VLFS, a surrogate model trained using the feed-forward neural network is proposed. The hydroelastic responses under different wavelengths, structural stiffnesses and wave directions are first generated where these data are split into three groups for training, validation, and testing (prediction) purposes. The accuracy of the prediction in terms of correlation coefficient R is compared for the different train datasets, the number of neurons and hidden layers as well as the optimisation techniques. The finding shows that an accuracy of close to 99% to the ground truth could be achieved with only 80% of the train dataset. The hydroelastic response under irregular wave conditions predicted using the feed-forward neural network framework is also presented. [ABSTRACT FROM AUTHOR]

Digital twins are undergoing growth that enables highly informative and scaleable interaction between physical objects and virtual twins, which is of great significance to the life cycle analysis of composites. Real-time fine evolution is challenging due to vast combinations of input features and high-resolution calculated variables. Here, we systematically demonstrate an AI-based methodology for digital twinning of complex composite structures. First, three types of deep neural networks are created with optionally used autoencoders as surrogate models, with architectures and data processing inspired by the rule-of-mixture of composites. Second, the prediction accuracy and efficiency are evaluated quantitatively and qualitatively, demonstrating the feasibility of predicting 3D displacement and stress fields directly from sensing data of temperature, pressure and loading displacement, and the optimal architecture is selected to be the evolving digital twin. Finally, the real-time interactive experiments are relayed to the digital twin and demonstrated that it can interact with physical objects and evolve online with high accuracy. These results indicate that the computational time can be reduced by 3∼6 orders of magnitude with high information intensity and scalability compared with conventional numerical and experimental methods, which opens up the avenues for the cost-effective and efficient development of digital twin services for composites. [Display omitted] [ABSTRACT FROM AUTHOR]

This paper presents a framework for integrating Explainable and Anomalous Machine Learning (EAML) into a digital twin to enable finetuning of mixtures as a mean to realize next-gen concretes with favorable performance. In this framework, both anomalous unsupervised and explainable supervised ML algorithms are joined to create a virtual assistant capable of exploring the influence of mixture materials and proportions on the required performance of concrete. This virtual assistant is not only trained to detect inherent vulnerabilities within mixtures but can also finetune such mixtures to overcome potential weaknesses – especially when concrete is expected to serve under extreme loading conditions. The proposed framework has been rigorously examined on three case studies to identify vulnerable mixtures to: 1) fire-induced spalling, 2) chloride penetration, and 3) failing to attain full design strength in job sites, using small and large datasets comprised from actual measurements. Results from our analysis show how the proposed framework was capable of identifying vulnerable concrete mixtures and of satisfying various performance metrics. While the proposed framework is designed to be algorithm-independent and hence can be scalable across multiple platforms, this work showcases the application of anomaly detecting and clustering algorithms, together with an ensemble of classifiers encompassing extreme and light gradient boosted trees (GBT), generalized additive models (GAM), and keras deep residual neural network (KDP). [ABSTRACT FROM AUTHOR]

• A fast and accurate fishing net damage detection model has been established based on digital twin. • Different damaged levels, positions and wave-current directions have been tested. • The digital twin model can accurately identify whether the fishing net is damaged or not. The fishing net is the most important component of aquaculture net cages. Once damaged, it can cause substantial economic losses and ecological problems to the aquaculture industry. To avoid these subsequent issues caused by the damage to the fishing net, seeking a computerized, labor-saving approach and detecting damage in real time becomes the primary task of this paper. Inspired by recent development of the artificial neural networks and machine learning, this study proposes a fast and accurate approach for fishing net damage detection based on digital twin. Firstly, time-domain numerical simulations of the fishing net are conducted in a series of wave and current conditions to develop the artificial neural network-based digital twin. Then, the significant wave height, the spectral peak period, and tensions of vertical and horizontal ropes are used as input variables during the artificial neural network training; the intact and damaged states of the fishing net are considered outputs. Alongside this, the back-propagation learning algorithm is used for training to maximize damage detection performance. The results highlight that the digital twin model can effectively identify the fishing net damage using the sensor data, and the accuracy of the damage detection reaches above 93%. When the damage occurs at different net positions, the prediction accuracies of the artificial neural network model for training and testing sets are 93.97% and 93.00%, respectively. Regarding different wave-current directions, the prediction accuracies of the artificial neural network model are 99.87 % and 99.34 % for training and testing sets. Moreover, the developed digital twin can accurately detect the damage of the net even sea conditions and sensor data are not included in the training. The digital twin model proposed in the present study can potentially be used for damage detection in other aquaculture components and structures. [ABSTRACT FROM AUTHOR]

This paper presents a method for detecting damage of a bottom-set gillnet based on sensor fusion and the Artificial Neural Network (ANN) model. In this regard, time-domain numerical simulations for a 300-m-long bottom-set gillnet with an equivalent-drag-net model are extensively performed. Various wave conditions as well as numerous damaged scenarios are considered in the numerical simulations, and extensive data are collected for the training and testing of the ANN-based machine-learning scheme. In training, representative sea states, net-assembly accelerations, and location-buoy displacements are selected as the input variables. The back-propagation learning algorithm is employed for training to maximize the damage-detection performance. The output of the ANN model is the identification of the type and location of the damaged net. The damage-detection capability is significantly enhanced by employing the moving standard deviation and median filter. The well-trained ANN models are shown to accurately (96% correct) detect the damage of the net even for the sea states not included in training. This study demonstrates that the real-time automatic monitoring of the underwater net systems can be designed by using the digital-twins technology with the applied ANN model and machine-learning scheme for the detection of damages and malfunctions. • The well-trained ANN models are shown to accurately (96% correct) detect the damage of the fishing net. • The number of neurons, median filter, and moving standard deviation can improve the performance of the ANN model. • The sensor-based fishing-net monitoring system can be designed by digital-twins technology with the applied ANN model. [ABSTRACT FROM AUTHOR]