Transfer Model Research Articles

Improved soil-to-plant transfer factors for 99Tc and 79Se in natural and agricultural ecosystems.

Geological Disposal Facilities (GDFs) are the preferred option for the disposal of high- and medium-level nuclear waste but environmental assessments for GDFs are complex. Models of transfer into the biosphere for radioisotopes that occur in nuclear waste rely on estimated Transfer Factors (TFs) that often have high levels of uncertainty and only exist for a few species. Here, using two key radioisotopes found in nuclear waste, we show that taxonomic analyses and phylogenetically based trait prediction (PTP) can be used to both reduce uncertainty in current estimates of soil-to-plant TF and to predict them for the many species with no measurements. We grew 61 species of plants selected to provide a phylogenetically informed sample, measured their uptake of 99Tc and 75Se, and reconstructed their possible evolutionary relationships using gene sequence information. The uptake of Tc and Se isotopes by plants was correlated, and for Tc was more similar within plant groups than between them and included significant taxonomic and phylogenetic influences. We use these findings to suggest improved soil-to-plant Transfer Factors (TFs) for99Tc and 79Se. We suggest that the approaches we used might be useful for a range of radionuclides, in both improving current estimates of TF and for predicting TFs to plants and, perhaps, to other biota. Such improvements might be useful not only for environmental assessments of nuclear waste disposal but also the environmental regulation of a nuclear industry being expanded in several nations to help meet targets for reducing global CO2 emissions.

Transferable machine learning approach for predicting electronic structures of charged defects

The study of electronic properties of charged defects plays a crucial role in advancing our understanding of how defects influence conductivity, magnetism, and optical behavior in various materials. However, despite its significance, research on large-scale defective systems has been hindered by the high computational cost associated with density functional theory (DFT). In this study, we propose HamGNN-Q, an E(3) equivariant graph neural network framework capable of accurately predicting DFT Hamiltonian matrices for diverse point defects with varying charges, utilizing a unified set of network weights. By incorporating background charge features into the element representation, HamGNN-Q facilitates a direct mapping from structural characteristics and background charges to the electronic Hamiltonian matrix of charged defect systems, obviating the need for DFT calculations. We showcase the model's high precision and transferability by evaluating its performance on GaAs systems encompassing diverse charged defect configurations. Furthermore, we predicted the wave function distribution of polarons induced by defects. We analyzed the node features through principal component analysis, providing physical insights for the interpretability of the HamGNN-Q model. Our approach provides a practical solution for accelerating electronic structure calculations of neutral and charged defects and advancing the design of materials with tailored electronic properties.

Transferability of the Surgical Attire Guideline Adherence Decision-Making Theory Beyond New England.

The AORN surgical attire guideline promotes cleanliness in the perioperative environment to minimize patients' risks of developing surgical site infections. In 2020, the surgical attire guideline adherence decision-making model was published based on findings from a study involving New England intraoperative team members. To explore the model's transferability across the United States, we replicated the 2020 study with intraoperative personnel who did not live or work in New England. The study results included the same core category as the 2020 study (ie, critical thinking), and we confirmed that intraoperative team members across the United States weigh the influence of various factors when deciding to what extent they would adhere to the surgical attire guideline. The results showed transferability of the theoretical model to perioperative settings across the United States. This model can be used to enhance intraoperative team members' surgical attire guideline adherence decisions.

Rapid Aerodynamic Prediction of Swept Wings via Physics-Embedded Transfer Learning

Machine-learning-based models provide a promising way to rapidly acquire transonic swept wing flow fields but suffer from large computational costs in establishing training datasets. Here, a physics-embedded transfer-learning framework is proposed to efficiently train the model by leveraging the idea that a three-dimensional flow field around wings can be analyzed with two-dimensional flow fields around cross-sectional airfoils. An airfoil aerodynamics prediction model is pretrained with airfoil samples. Then, an airfoil-to-wing transfer model is fine-tuned with a few wing samples to predict three-dimensional flow fields based on two-dimensional results on each spanwise cross section. Sweep theory is embedded when determining the corresponding airfoil geometry and operating conditions, and to obtain the sectional airfoil lift coefficient, which is one of the operating conditions, the low-fidelity vortex lattice method and data-driven methods are proposed and evaluated. Compared to a nontransfer model, introducing the pretrained model reduces the error by 30%, whereas introducing sweep theory further reduces the error by 9%. When reducing the dataset size, less than half of the wing training samples are need to reach the same error level as the nontransfer framework, which makes establishing the model much easier.

Abnormal-Sound Diagnosis for Kaplan Hydroelectric Generating Units Based on Continuous Wavelet Transform and Transfer Learning.

To realize abnormal-sound diagnosis in hydroelectric generating units, this study proposes a method based on continuous wavelet transform (CWT) and Transfer Learning (TL). A denoising algorithm utilizing spectral noise-gate technology is proposed to enhance fault characteristics in hydroelectric units. Subsequently, Continuous Wavelet Transform is applied to obtain frequency components, and the results are converted into a series of pseudo-color images to highlight information differences. A transfer model is subsequently developed for feature extraction, utilizing simplified fully connected layers to reduce modeling costs. The study optimizes key parameters during the signal-processing stage and achieves an improved parameter-setting scheme. Acoustic signals corresponding to four different fault states and a normal state are collected from a Kaplan hydroelectric generating unit in a hydropower station. The signal diagnosis accuracy rates before filtering are 84.83% and 95.14%. These rates significantly improved to 98.88% and 98.06%, respectively, demonstrating the effectiveness of the noise-reduction process. To demonstrate the superiority of the improved model in this work, a series of classic deep-learning models, including AlexNet, Resnet18, and MobileNetV3, are used for comparative analysis. The proposed method can effectively diagnose faults in Kaplan hydroelectric generating units with a high accuracy, which is crucial for the daily monitoring and maintenance of these units.

Kinetic modeling and optimization of biogas production from food waste and cow manure co-digestion

As global efforts to reduce reliance on fossil fuels and improve waste management practices intensify, the development of efficient biogas production methods offers a promising solution. This study addresses the urgent need for renewable energy and effective waste management by investigating biogas production from the anaerobic digestion of food waste and cow manure. Optimal conditions for mono-digestion of cow manure were identified at 8 % total solid in 38 °C. In co-digestion experiments, varying food waste ratios were tested, with a 6:2 ratio based on total solid achieving the highest biogas yield of 325 mL/g VS. The synergistic effects of co-digestion led to enhanced nutrient balance and biogas production, significantly outperforming mono-digestion. The experimental data were analyzed using first-order kinetics, modified Gompertz, modified logistic, and transference models, with the modified Gompertz model providing the best fit (R²=0.999). Results showed improved biogas production and higher rates of volatile solids and COD removal. This study focuses on maintaining a constant optimized total solids concentration for optimizing feedstock composition to enhance biogas production and compares four kinetic models on predicting performance, providing insights into anaerobic digestion system improvement. The study's limitations, such as the lack of post-optimization TS adjustments, absence of pH control, fixed temperature conditions, highlight areas for future work to provide more insights and improve co-digestion efficiency.

Text Summarization System by Deep Learning & Transfer Model

This paper evaluates recent advances in text summarization strategies, focusing at the strategies developed and deployed in a practical assignment. The project makes use of both abstractive and extractive summarization methods, employing deep gaining knowledge of models such as the Transformer architecture. This paper will discover the comparative overall performance of various models, discuss their deployment, and highlight destiny potentialities for boosting textual content summarization technology.

Anaerobic digestion performance and kinetics of biomass pretreated with various fungal strains utilizing exponential and sigmoidal equation models

Corn silage was pretreated with white rot fungi to optimize the methane generation capacity of biomass and the effect of the most important pretreatment parameters on the anaerobic digestion performance of material was determined by employing mathematical modelling. Pretreatment with selective fungi led to optimal increment in methane production from 0.301 m3kgVS−1 from corn silage to 0.465 m3kgVS−1 from the 10-day modified material, whereas non-selective fungi had negative effect. A reverse L-shape curve observed for the corn silage indicated an uninterrupted methane production and was not modified when the substrate was pretreated with Dichomitus squalens, Irpex lacteus and Trametes versicolor. Exponential models fitted L-curves better than sigmoidal equations with Cone model providing the most accurate predictions regardless of the selective or non-selective feature of the fungal strain used for the pretreatment of corn silage. Increasing pretreatment time led to a differentiation of the methane yield and adjustment of the curve shape to a stepped outline adequately modelled by Cone and transference models. Contrary to three-parameter sigmoidal models a fourth shape variable included in Richards equation was crucial, allowing for a flexible inflection point that enabled an accurate replication of an S-shape curve obtained for methane production from the hybrid substrate.

Construction frontier molecular orbital prediction model with transfer learning for organic materials

The frontier molecular orbitals of organic semiconductor materials play a crucial role in the performance of photoelectric devices, including organic photovoltaics (OPVs), organic light-emitting diodes (OLEDs), and organic photodetectors (OPDs). In this work, a model for predicting frontier molecular orbital of organic materials, including HOMO and LUMO levels, is established with the extreme gradient boosting algorithm and Klekota-Roth fingerprints. The correlation coefficients of HOMO or LUMO energy levels in the testing set are 0.75 and 0.84 in the transfer model from 11,626 DFT data in Harvard Energy database to 1198 experimental data in literature. The difference between the ML predicted value and the experimental value is smaller than the difference between ML prediction and DFT calculation, always less than 10%. Moreover, based on correlation and SHAP interpretability analysis, 13 key structural fragments influencing energy levels are selected to further verify the effective regulation of the frontier molecular orbital by the key structural fragments in practical applications. Considering the completely opposite regulatory functions of key structural fragments on HOMO and LUMO energy levels, four new Y6 derivatives, Y-PCP, Y-P6F, Y-PCF, and Y-P4FC, are designed to flexibly modify the HOMO and LUMO energy levels. The prediction trends of ML align closely with the computational trends from DFT. It is worth noting that the accuracy of LUMO energy level prediction by the prediction model makes up for the instability of DFT calculation on LUMO energy level. This work offers a cost-effective method to accelerate the acquisition of electronic properties of organic materials.

Biomethane production modelling from third-generation biomass

Third-generation biomass constitutes a renewable energy source that could substitute fossil fuels. This study evaluated the biomethane potential (BMP) of Ulva sp., Codium sp. and Undaria pinnatifida through anaerobic digestion. The daily production of biomethane was evaluated using different models, including Modified Gompertz, Chen and Hashimoto, First-order, Transfer and Cone models, as well as an Artificial Neural Network (ANN) model. The experimental BMP was 0.17, 0.26, and 0.32 Nm³/kg VS for Ulva sp., Codium sp., and U pinnatifida, respectively. Among the non-linear regression models, the Transfer model (R2 > 0.9915) and the First-order model (R2 > 0.9889) are the ones that best fit the experimental data. However, the ANN shows a better fit (R2 > 0.999 and RMSE<4.537) to the data compared to the non-linear regression models. Furthermore, ANN can capture the complexity of biological systems, allowing for more accurate and detailed modelling of the processes involved. The identification and optimization of the biomethane potential of macroalgae contribute to developing sustainable energy alternatives, offering a renewable energy source that could mitigate the environmental stress associated with traditional fossil fuels.

The application of direct ridership models in the evaluation of the expansion of the Porto Light Rail Transit

The main purpose of this paper is to show how a direct demand ride model coupled with a transfer model was able to support the choice of alternative routes for the expansion of the light rail system serving the Porto Metropolitan Area. After an overview of the literature on direct ridership models, emphasizing some key issues such as the need for a systematic assessment of their forecasting performance, the issues related to the definition of the pedestrian catchment area and the limitations of the simultaneous consideration of demand and supply effects, the paper moves into the case study, providing some background information on current occupation densities, land uses and mobility patterns, as well as on the performance of the existing LRT in the Metropolitan Area of Porto. The development of the direct ridership model, measuring the potential attractiveness of each station, and the transfer model, measuring the number of transfers at each station, are presented in detail. A justification is provided why, in this case, a two-step modelling approach was necessary. Further details about the statistical tests for model validation are also provided. After a brief characterization of the alternative routes under analysis for the expansion of the network, the modelling results are presented enabling a comparative assessment of the potential performance of each proposed route. The paper ends with a discussion of the relevance of this modelling results vis a vis the actual final decisions on investment priorities taken jointly by the Metropolitan Council and the Metro Company.

Constructing a temperature transferable coarse-grained model of cis-1,4-polyisoprene with the structural and thermodynamic consistency aided by machine learning

Polyisoprene (PI) is a widely used polymer and constructing a systematic coarse-grained (CG) PI model with the structural and thermodynamic consistency with the underlying atomic model over a wide range of thermodynamic conditions is very important for the predictive capability of CG model on overall properties of PI polymer materials and the establishment of their structure-property relationship. However, as the number of tunable CG potential parameters and target properties grows, traditional parameter tuning methods become impractical. In this work, we present a novel approach for determining the optimal CGPI non-bonded potential parameters by employing Particle Swarm Optimization as the calibrator with machine learning-based models trained using molecular dynamics data. The resulting CG model is further augmented with temperature factors through a multistate parameterization approach. This enhancement ensures the model's temperature transferability of structure and thermodynamics in a wide temperature of 150K∼750K.

Increasing drought frequency in the central Zagros Mountains of western Iran over the past two centuries

The Zagros region in western Iran has experienced prolonged drought, significantly impacting water resource and forest ecosystems over the past few centuries. Understanding historical prolonged drought is crucial for making precise forecasts of shifts in regional drought in the Zagros region. Due to the lack of comprehensive historical records, the use of proxy records is considered a valuable tool for reconstructing past drought variations. We aimed to construct a tree-ring chronology of Juniper (Juniperus polycarpus) in the central Zagros region to comprehend its growth response to climate variables. We cored 25 J. polycarpus trees from the Keygooran forest reserve in western Iran and developed the tree ring chronology (1802–2022) using dplR. The relationships between tree growth and climate variables of monthly mean temperature, precipitation, PDSI (Palmer Drought Severity Index), and SPEI (Standardized Precipitation Evapotranspiration Index) were examined. The analysis revealed a strong positive correlation between tree growth and SPEI March-September of 1990–2018 on a 48-month timescale with a 2-year lag (r = 0.61, p < 0.05) which was included in the transfer model. Consequently, the SPEI-48 March-September was reconstructed for the period 1802–2022. On this reconstruction, several extremely dry years in 1832, 1867, and 1876 and extremely wet years in 1807, 1814, 1838, 1850, 1907, and 1908 years were identified, some of them aligning with historical records in Iran. Furthermore, there was a relationship between SPEI-48 March-September and teleconnection events suggesting that certain drought occurrences in the area were associated with the El Niño-Southern Oscillation (ENSO). Consequently, this reconstruction can serve as a reliable proxy for large-scale drought variability in the Zagros region of western Iran. Moreover, our research underscores the utility of SPEI in reconstructing the history of semi-arid climate conditions.

SegmentAnyTree: A sensor and platform agnostic deep learning model for tree segmentation using laser scanning data

This study focuses on advancing individual tree crown (ITC) segmentation in lidar data, developing a sensor- and platform-agnostic deep learning model transferable across a spectrum of dense laser scanning datasets from drone (ULS), to terrestrial (TLS), and mobile (MLS) laser scanning data. In a field where transferability across different data characteristics has been a longstanding challenge, this research marks a step towards versatile, efficient, and comprehensive 3D forest scene analysis.Central to this study is model performance evaluation based on platform type (ULS vs. MLS) and data density. This involved five distinct scenarios, each integrating different combinations of input training data, including ULS, MLS, and their augmented versions through random subsampling, to assess the model's transferability to varying resolutions and efficacy across different canopy layers. The core of the model, inspired by the PointGroup architecture, is a 3D convolutional neural network (CNN) with dedicated prediction heads for semantic and instance segmentation. The model underwent comprehensive validation on publicly available, machine learning-ready point cloud datasets. Additional analyses assessed model adaptability to different resolutions and performance across canopy layers.Our results reveal that point cloud random subsampling is an effective augmentation strategy and improves model performance and transferability. The model trained using the most aggressive augmentation, including point clouds as sparse as 10 points m−2, showed best performance and was found to be transferable to sparse lidar data and boosts detection and segmentation of codominant and dominated trees. Notably, the model showed consistent performance for point clouds with densities >50 points m−2 but exhibited a drop in performance at the sparsest level (10 points m−2), mainly due to increased omission rates. Benchmarking against current state-of-the-art methods revealed boosts of up to 20% in the detection rates, indicating the model's superior performance on multiple open benchmark datasets. Further, our experiments also set new performance baselines for the other public datasets. The comparison highlights the model's superior segmentation skill, mainly due to better detection and segmentation of understory trees below the canopy, with reduced computational demands compared to other recent methods.In conclusion, the present study demonstrates that it is indeed feasible to train a sensor-agnostic model that can handle diverse laser scanning data, going beyond current sensor-specific methodologies. Further, our study sets a new baseline for tree segmentation, especially in complex forest structures. By advancing the state-of-the-art in forest lidar analysis, our work also lays the foundation for future innovations in ecological modeling and forest management.

AlphaFold 2-based stacking model for protein solubility prediction and its transferability on seed storage proteins

Accurate protein solubility prediction is crucial in screening suitable candidates for food application. Existing models often rely only on sequences, overlooking important structural details. In this study, a regression model for protein solubility was developed using both the sequences and predicted structures of 2983 E. coli proteins. The sequence and structural level properties of the proteins were bioinformatically extracted and subjected to multilayer perceptron (MLP). Moreover, residue level features and contact maps were utilized to construct a graph convolutional network (GCN). The out-of-fold predictions of the two models were combined and fed into multiple meta-regressors to create a stacking model. The stacking model with support vector regressor (SVR) achieved R2 of 0.502 and 0.468 on test and external validation datasets, respectively, displaying higher performance compared to existing regression models. Based on the improved performance compared to its based models, the stacking model effectively captured the strength of its base models as well as the significance of the different features used. Furthermore, the model's transferability was indirectly validated on a dataset of seed storage proteins using Osborne definition as well as on a case study using molecular dynamic simulation, showing potential for application beyond microbial proteins to food and agriculture-related ones.

Signal enhancement method for gearboxes fault diagnosis in robotic flexible joint

Abstract Motor Current Signal Analysis (MCSA) provides a non-intrusive approach to fault diagnosis. However, the fault impact reacting in the current is reduced due to the presence of flexible structures in the transmission path from the fault source to the motor. Therefore, this paper proposes a method to enhance the frequency domain of the current signal of a single mechanical fault through a transfer model between motor torque and link vibration. First, the joint system dynamics model was developed based on a three-inertia simplified model. The transfer model of motor torque and link vibration was defined based on the system dynamics. The link vibration is then estimated based on the transfer model and electromagnetic torque. Link vibration signal is considered as an enhancement of the torque signal. Finally, the link vibration signature analysis is performed instead of MCSA. The experimental results show that the method is effective in enhancing the features of individual mechanical faults and improving the fault diagnosis performance.

Cross-machine predictions of the quality of injection-molded parts by combining machine learning, quality indices, and a transfer model

Cross-machine predictions of the quality of injection-molded parts by combining machine learning, quality indices, and a transfer model

Identification of metastasis-related genes for predicting prostate cancer diagnosis, metastasis and immunotherapy drug candidates using machine learning approaches

BackgroundProstate cancer (PCa) is the second leading cause of tumor-related mortality in men. Metastasis from advanced tumors is the primary cause of death among patients. Identifying novel and effective biomarkers is essential for understanding the mechanisms of metastasis in PCa patients and developing successful interventions.MethodsUsing the GSE8511 and GSE27616 data sets, 21 metastasis-related genes were identified through the weighted gene co-expression network analysis (WGCNA) method. Subsequent functional analysis of these genes was conducted on the gene set cancer analysis (GSCA) website. Cluster analysis was utilized to explore the relationship between these genes, immune infiltration in PCa, and the efficacy of targeted drug IC50 scores. Machine learning algorithms were then employed to construct diagnostic and prognostic models, assessing their predictive accuracy. Additionally, multivariate COX regression analysis highlighted the significant role of POLD1 and examined its association with DNA methylation. Finally, molecular docking and immunohistochemistry experiments were carried out to assess the binding affinity of POLD1 to PCa drugs and its impact on PCa prognosis.ResultsThe study identified 21 metastasis-related genes using the WGCNA method, which were found to be associated with DNA damage, hormone AR activation, and inhibition of the RTK pathway. Cluster analysis confirmed a significant correlation between these genes and PCa metastasis, particularly in the context of immunotherapy and targeted therapy drugs. A diagnostic model combining multiple machine learning algorithms showed strong predictive capabilities for PCa diagnosis, while a transfer model using the LASSO algorithm also yielded promising results. POLD1 emerged as a key prognostic gene among the metastatic genes, showing associations with DNA methylation. Molecular docking experiments supported its high affinity with PCa-targeted drugs. Immunohistochemistry experiments further validated that increased POLD1 expression is linked to poor prognosis in PCa patients.ConclusionsThe developed diagnostic and metastasis models provide substantial value for patients with prostate cancer. The discovery of POLD1 as a novel biomarker related to prostate cancer metastasis offers a promising avenue for enhancing treatment of prostate cancer metastasis.

Innovative Design and Experimental Research of the Patient Transfer Apparatus for MRI Room

Background: Patient Transfer Apparatus (PTA), which has been reported by various relevant papers and patents, is widely used in the hospital. However, there are few corresponding transfer apparatuses for research and development in the nuclear magnetic resonance imaging (MRI) room because of the influence of a high magnetic field environment. It is desired to require the apparatus for the patient and the medical staff to have weak magnetism, high matching, and easy operation, etc. Objective: The purpose of this study is to find out the working principle of PTA, to conduct parameter optimization and design and develop more effective PTA for the application of the MRI room. Methods: Firstly, based on the patient transfer process, a novel transfer model with the coordinated movement of the transfer belt and the moving panel is proposed, the corresponding clutch mechanism is conceived and the working principle of the whole mechanism movement is designed and analyzed in detail. Secondly, the force analysis of the clutch mechanism, lifting mechanism and patient transfer mechanism are performed, the mechanical structures are optimized, and the optimal sizes are obtained. Thirdly, the mechanical structure of the system is designed in detail, and the prototype is manufactured. Results: Finally, the performance evaluation of the system is conducted by means of the fuzzy evaluation method and clinical study. The results showed that the motion function of the apparatus is reasonable and it can work normally in the MRI room, the comfort of the PTA is also excellent. The research results also prove the accuracy of the working principle and the rationality of the structural design. Conclusion: The non-magnetic PTA is suitable for quickly transferring patients who are unable to move in the MRI room of a hospital by the manual method.

A knowledge-aware deep learning model for landslide susceptibility assessment in Hong Kong

Despite the success of the growing data-driven landslide susceptibility prediction, the model training heavily relies on the quality of the data (involving topography, geology, hydrology, land cover, climate, and human activity), the structure of the model, and the fine-tuning of the model parameters. Few data-driven methods have considered incorporating ‘landslide priors’, as in this article the prior knowledge or statistics related to landslide occurrence, to enhance the model's perception in landslide mechanism. The main objective and contribution of this study is the coupling of landslide priors and a deep learning model to improve the model's transferability and stability. This is accomplished by selecting non-landslide sample grounded on landslide statistics, disentangling input landslide features using a variational autoencoder, and crafting a loss function with physical constraints. This study utilizes the SHAP method to interpret the deep learning model, aiding in the acquisition of feature permutation results to identify underlying landslide causes. The interpretation result indicates that ‘slope’ is the most influential factor. Considering the extreme rainfall impact on landslide occurrences in Hong Kong, we combine this prior into the deep learning model and find feature ranking for ‘rainfall’ improved, in comparison to the ranking result interpreted from a pure MLP. Further, the potency of MT-InSAR is utilized to augment the landslide susceptibility map and promote efficient cross-validation. A comparison of InSAR results with historical images reveals that detectable movement before their occurrence is evident in only a minority of landslides. Most landslides occur spontaneously, exhibiting no precursor motion. Comparing with other data-driven methods, the proposed methods outperform in accuracy (by 2 %–5 %), precision (by 2 %–7 %), recall (by 1 %–3 %), F1-score (by 8 %–10 %), and AuROC (by 2 %–4 %). Especially, the Cohen Kappa performance surpasses nearly 20 %, indicating that the knowledge-aware methodology enhances model generalization and mitigates training bias induced by unbalanced positive and negative samples.