2063 research outputs found
Sort by
Diagnosis of the health status of mooring systems for floating offshore wind turbines using autoencoders
Get PDFFloating offshore wind turbines (FOWTs) show promise in terms of energy production, availability, and sustainability, but remain unprofitable due to high maintenance costs. This work proposes a deep learning algorithm to detect mooring line degradation and failure by monitoring the dynamic response of the publicly available DeepCWind OC4 semi-submersible platform. This study implements an autoencoder capable of predicting multiple forms of damage occurring at once, with various levels of severity. Given the scarcity of real data, simulations performed in OpenFAST, recreating both healthy and damaged mooring systems, are used to train and validate the algorithm. The novelty of the proposed approach consists of using a set of key statistical metrics describing the platform’s displacements and rotations as input layer for the autoencoder. The statistics of the responses are calculated at 33-minute-long sea states under a broad spectrum of metocean and wind conditions. An autoencoder is trained using these parameters to discover that the proposed algorithm is capable of detecting mild anomalies caused by biofouling and anchor displacements, with correlation coefficients up to 98.51\% and 99.16\%, respectively. These results are encouraging for the continuous health monitoring of FOWT mooring systems using easily measurable quantities to plan preventive maintenance actions adequately
Derivative curve estimation in longitudinal studies using P-splines
Get PDFThe estimation of curve derivatives is of interest in many disciplines. It allows the extraction of important characteristics to gain insight about the underlying process. In the context of longitudinal data, the derivative allows the description of biological features of the individuals or finding change regions of interest. Although there are several approaches to estimate subject-specific curves and their derivatives, there are still open problems due to the complicated nature of these time course processes. In this article, we illustrate the use of P-spline models to estimate derivatives in the context of longitudinal data. We also propose a new penalty acting at the population and the subject-specific levels to address under-smoothing and boundary problems in derivative estimation. The practical performance of the proposal is evaluated through simulations, and comparisons with an alternative method are reported.
Finally, an application to longitudinal height measurements of 125 football players in a youth professional academy is presented, where the goal is to analyse their growth and maturity patterns over time.RYC2019-027534-I
The Medical Services of Athletic Clu
Neural network architecture optimization using automated machine learning for borehole resistivity measurements
Get PDFDeep neural networks (DNNs) offer a real-time solution for the inversion of borehole resistivity measurements to approximate forward and inverse operators. Using extremely large DNNs to approximate the operators is possible, but it demands considerable training time. Moreover, evaluating the network after training also requires a significant amount of memory and processing power. In addition, we may overfit the model. In this work, we propose a scoring function that accounts for the accuracy and size of the DNNs compared to a reference DNNs that provides good approximations for the operators. Using this scoring function, we use DNN architecture search algorithms to obtain a quasi-optimal DNN smaller than the reference network; hence, it requires less computational effort during training and evaluation. The quasi-optimal DNN delivers comparable accuracy to the original large DNN.PDC2021-121093-I00
IA4TES
RYC2021-032853-
Synchronization in STDP-driven memristive neural networks with time-varying topology
Get PDFSynchronization is a widespread phenomenon in the brain. Despite numerous studies, the specific parameter configurations of the synaptic network structure and learning rules needed to achieve robust and enduring synchronization in neurons driven by spike-timing-dependent plasticity (STDP) and temporal networks subject to homeostatic structural plasticity (HSP) rules remain unclear. Here, we bridge this gap by determining the configurations required to achieve high and stable degrees of complete synchronization (CS) and phase synchronization (PS) in time-varying small-world and random neural networks driven by STDP and HSP. In particular, we found that decreasing P (which enhances the strengthening effect of STDP on the average synaptic weight) and increasing F (which speeds up the swapping rate of synapses between neurons) always lead to higher and more stable degrees of CS and PS in small-world and random networks, provided that the network parameters such as the synaptic time delay τc , the average degree ⟨ k⟩ , and the rewiring probability β have some appropriate values. When τc , ⟨ k⟩ , and β are not fixed at these appropriate values, the degree and stability of CS and PS may increase or decrease when F increases, depending on the network topology. It is also found that the time delay τc can induce intermittent CS and PS whose occurrence is independent F. Our results could have applications in designing neuromorphic circuits for optimal information processing and transmission via synchronization phenomena.Deutsche Forschungsgemeinschaft (DFG, German Research Foundation) – Project No. 456989199
Elkartek 2023 via grant ONBODY no. KK-2023/00070
Validating the effect of fuel moisture content by a multivalued operator in a simplified physical fire spread model
Get PDFFuel moisture content (FMC) plays a significant role in wildfire behavior and rate of spread (ROS). In addition, FMC is a highly dynamic factor and very vulnerable to climate variations. Understanding the effect of FMC on the behavior of fire spread models is crucial, and detailed analysis of specific aspects of complex models is a very effective way to improve them. The simplified physical fire spread model PhyFire considers the effect of FMC in a novel way, involving a multivalued maximal monotone operator. Several numerical experiments have been carried out to confirm that the behavior of the ROS simulated with PhyFire involving FMC is as expected in the reviewed literature: an exponential decrease in fire ROS compared to FMC, for different scenarios, considering different fuel types, terrain slopes and wind speeds. PhyFire performs very accurately, proving that the multivalued operator used is suitable and consistent
Role of immigration and emigration on the spread of COVID-19 in a multipatch environment: a case study of India
No full textHuman mobility has played a critical role in the spread of COVID-19. The understanding of mobility helps in getting information on the acceleration or control of the spread of disease. The COVID-19 virus has been spreading among several locations despite all the best efforts related to its isolation. To comprehend this, a multi-patch mathematical model of COVID-19 is proposed and analysed in this work, where-in limited medical resources, quarantining, and inhibitory behaviour of healthy individuals are incorporated into the model. Furthermore, as an example, the impact of mobility in a three-patch model is studied considering the three worst-hit states of India, i.e. Kerala, Maharashtra and Tamil Nadu, as three patches. Key parameters and the basic reproduction number are estimated from the available data. Through results and analyses, it is seen that Kerala has a higher effective contact rate and has the highest prevalence. Moreover, if Kerala is isolated from Maharashtra or Tamil Nadu, the number of active cases will increase in Kerala but reduce in the other two states. Our findings indicate that the number of active cases will decrease in the high prevalence state and increase in the lower prevalence states if the emigration rate is higher than the immigration rate in the high prevalence state. Overall, proper travel restrictions are to be implemented to reduce or control the spread of disease from the high-prevalence state to other states with lower prevalence rates
Decentralized Federated Learning for Epileptic Seizures Detection in Low-Power Wearable Systems
Get PDFIn healthcare, data privacy of patients regulations prohibits data from being moved outside the hospital, preventing international medical datasets from being centralized for AI training. Federated learning (FL) is a data privacy-focused method that trains a global model by aggregating local models from hospitals. Existing FL techniques adopt a central server-based network topology, where the server assembles the local models trained in each hospital to create a global model. However, the server could be a point of failure, and models trained in FL usually have worse performance than those trained in the centralized learning manner when the patient's data are not independent and identically distributed (Non-IID) in the hospitals. This paper presents a decentralized FL framework, including training with adaptive ensemble learning and a deployment phase using knowledge distillation. The adaptive ensemble learning step in the training phase leads to the acquisition of a specific model for each hospital that is the optimal combination of local models and models from other available hospitals. This step solves the non-IID challenges in each hospital. The deployment phase adjusts the model's complexity to meet the resource constraints of wearable systems. We evaluated the performance of our approach on edge computing platforms using EPILEPSIAE and TUSZ databases, which are public epilepsy datasets.RYC2021-032853-
Impact of H2/CH4 blends on the flexibility of micromix burners applied to industrial combustion systems
Get PDFThe present paper investigates the feasibility of using H2/CH4 fuel blends in micromix-type burners applied to industrial combustion systems. The micromix burner concept, characterised by the formation of miniaturised and compact turbulent diffusion flames, was developed for gas turbine hydrogen burners showing low NOx emissions (below 10 ppm) without flashback risk, which represent the main issues when using pure hydrogen or hydrogen enriched natural gas blends as fuel, making it a promising concept to be applied in industrial burners. The study was carried out through numerical CFD simulations, accounting for detailed chemistry calculations of turbulent micromix flames and previously validated through experimental measurements in a laboratory-scale micromix burner prototype. The resulting flow, temperature and exhaust emission characteristics for three H2/CH4 fuel blends with H2 content of 90, 75 and 60% respectively were analysed and discussed for air–fuel equivalence ratios at λ=1.8 and 1.6 (lower than the well-characterised air–fuel equivalence ratios in micromix gas turbine burners at λ=2.5 and closer to current industrial burners), considering an energy density of 14 MW/m2 bar. Numerical results showed low fuel flexibility for industrial-scale micromix burners, with still low NOx emissions (12–85 ppm) but relatively high CO emissions (448–4970 ppm) for the considered blends and λ values. The lowest CO emissions were given together with jet penetration phenomena, ruling out the feasibility of these design points due to the greater importance of the latter phenomenon.Provincial Council of Bizkaia within the Technology
Transfer Programme 2021,co-financed by EGEF, Basque Governmen
Clinical prediction rules for adverse evolution in patients with COVID-19 by the Omicron variant
Get PDFObjective: We identify factors related to SARS-CoV-2 infection linked to hospitalization, ICU admission, and mortality and develop clinical prediction rules.
Methods: Retrospective cohort study of 380,081 patients with SARS-CoV-2 infection from March 1, 2020 to January 9, 2022, including a subsample of 46,402 patients who attended Emergency Departments (EDs) having data on vital signs. For derivation and external validation of the prediction rule, two different periods were considered: before and after emergence of the Omicron variant, respectively. Data collected included sociodemographic data, COVID-19 vaccination status, baseline comorbidities and treatments, other background data and vital signs at triage at EDs. The predictive models for the EDs and the whole samples were developed using multivariate logistic regression models using Lasso penalization.
Results: In the multivariable models, common predictive factors of death among EDs patients were greater age; being male; having no vaccination, dementia; heart failure; liver and kidney disease; hemiplegia or paraplegia; coagulopathy; interstitial pulmonary disease; malignant tumors; use chronic systemic use of steroids, higher temperature, low O2 saturation and altered blood pressure-heart rate. The predictors of an adverse evolution were the same, with the exception of liver disease and the inclusion of cystic fibrosis. Similar predictors were found to be related to hospital admission, including liver disease, arterial hypertension, and basal prescription of immunosuppressants. Similarly, models for the whole sample, without vital signs, are presented.
Conclusions: We propose risk scales, based on basic information, easily-calculable, high-predictive that also function with the current Omicron variant and may help manage such patients in primary, emergency, and hospital care.
Keywords: COVID-19; Clinical decision rules; Health care; Outcome assessment; SARS-CoV-2
Numerical quadrature for Gregory quads
Get PDFWe investigate quadrature rules in the context of quadrilateral Gregory patches, in short Gregory quads. We provide numerical and where possible symbolic quadrature rules for the space spanned by the twenty polynomial/rational functions associated with Gregory quads, as well as the derived spaces including derivatives, products, and products of derivatives of these functions. This opens up the possibility for a wider adoption of Gregory quads in numerical simulations