Initial Dataset Research Articles

The Effects of Genetic Distance and Genetic Diversity on Genomic Prediction Accuracy for Soybean Quantitative Disease Resistance to Phytophthora sojae

Assessing quantitative disease resistance and other complex traits can be time- and space-consuming, limiting the number of lines that can be evaluated in a breeding program. A potential improvement is the application of genomic prediction models, where genotypic data are used to calculate genomic estimated breeding values, referred to as genomic selection. Genomic prediction uses training datasets, where a germplasm panel is phenotyped and genotyped to calculate genomic estimated breeding values in a validation panel of lines based solely on genotypic data. To develop an initial phenotypic dataset, breeders may consider utilizing previously phenotyped lines in a publicly available dataset, such as plant introductions (PIs) from the USDA Soybean collection. A relevant question is, how effective is genomic prediction across diverse training and prediction panels? To answer this, we used previously collected phenotypic data for quantitative disease resistance toward Phytophthora sojae. Diverse germplasm panels were represented by sets of PIs originating from the United States, the Republic of Korea, and worldwide, for a total collection of 1,768 PIs. The accuracy of the prediction model was significantly influenced by population differentiation between panels, as well as genetic and phenotypic diversity. Low prediction accuracy resulted when the panel used to create the prediction model was highly differentiated from the validation panel. However, the addition of a small number of accessions to the training panel that were more closely related to the validation panel resulted in increased accuracy. [Formula: see text] Copyright © 2024 The Author(s). This is an open access article distributed under the CC BY-NC-ND 4.0 International license .

Systematic comparison of 3D Deep learning and classical machine learning explanations for Alzheimer’s Disease detection

Black-box deep learning (DL) models trained for the early detection of Alzheimer’s Disease (AD) often lack systematic model interpretation. This work computes the activated brain regions during DL and compares those with classical Machine Learning (ML) explanations. The architectures used for DL were 3D DenseNets, EfficientNets, and Squeeze-and-Excitation (SE) networks. The classical models include Random Forests (RFs), Support Vector Machines (SVMs), eXtreme Gradient Boosting (XGBoost), Light Gradient Boosting (LightGBM), Decision Trees (DTs), and Logistic Regression (LR). For explanations, SHapley Additive exPlanations (SHAP) values, Local Interpretable Model-agnostic Explanations (LIME), Gradient-weighted Class Activation Mapping (GradCAM), GradCAM++ and permutation-based feature importance were implemented. During interpretation, correlated features were consolidated into aspects. All models were trained on the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. The validation includes internal and external validation on the Australian Imaging and Lifestyle flagship study of Ageing (AIBL) and the Open Access Series of Imaging Studies (OASIS).DL and ML models reached similar classification performances. Regarding the brain regions, both types focus on different regions. The ML models focus on the inferior and middle temporal gyri, and the hippocampus, and amygdala regions previously associated with AD. The DL models focus on a wider range of regions including the optical chiasm, the entorhinal cortices, the left and right vessels, and the 4th ventricle which were partially associated with AD. One explanation for the differences is the input features (textures vs. volumes). Both types show reasonable similarity to a ground truth Voxel-Based Morphometry (VBM) analysis. Slightly higher similarities were measured for ML models.

Positive Mass Theorems for Spin Initial Data Sets With Arbitrary Ends and Dominant Energy Shields

Abstract We prove a positive mass theorem for spin initial data sets $(M,g,k)$ that contain an asymptotically flat end and a shield of dominant energy (a subset of $M$ on which the dominant energy scalar $\mu -|J|$ has a positive lower bound). In a similar vein, we show that for an asymptotically flat end $\mathcal{E}$ that violates the positive mass theorem (i.e., $\textrm{E} < |\textrm{P}|$), there exists a constant $R>0$, depending only on $\mathcal{E}$, such that any initial data set containing $\mathcal{E}$ must violate the hypotheses of Witten’s proof of the positive mass theorem in an $R$-neighborhood of $\mathcal{E}$. This implies the positive mass theorem for spin initial data sets with arbitrary ends, and we also prove a rigidity statement. Our proofs are based on a modification of Witten’s approach to the positive mass theorem involving an additional independent timelike direction in the spinor bundle.

Personalized Prediction of Parkinson's Disease Progression Based on Deep Gaussian Processes.

Parkinson's disease is a chronic progressive neurodegenerative disease with highly heterogeneous symptoms and progression. It is helpful for patient management to establish a personalized model that integrates heterogeneous interpretation methods to predict disease progression. In the study, we propose a novel approach based on a multi-task learning framework to divide Parkinson's disease progression modeling into an unsupervised clustering task and a disease progression prediction task. On the one hand, the method can cluster patients with different progression trajectories and discover new progression patterns of Parkinson's disease. On the other hand, the discovery of new progression patterns helps to predict the future progression of Parkinson's disease markers more accurately through parameter sharing among multiple tasks. We discovered three different Parkinson's disease progression patterns and achieved better prediction performance (MAE=5.015, RMSE=7.284, r2=0.727) than previously proposed methods on Parkinson's Progression Markers Initiative datasets, which is a longitudinal cohort study with newly diagnosed Parkinson's disease.

DATA SCIENCE: Data Visualization and Data Analytics in the Process of Data Mining

In the rapidly evolving landscape of data mining, the effective extraction of valuable insights from large datasets is paramount. This survey paper investigates the pivotal roles of data visualization and analytics in the intricate process of data mining abstraction. We delve into the symbiotic relationship between these two components, examining how they synergistically contribute to the extraction, representation, and interpretation of meaningful patterns and trends within complex datasets. The survey begins by elucidating the fundamental concepts of data mining abstraction and the significance of distilling actionable knowledge from raw data. It subsequently explores the multifaceted benefits of data visualization, elucidating its role in pattern identification, insight generation, and seamless communication of findings to diverse stakeholders.In parallel, the paper navigates through the landscape of data analytics, unraveling its diverse methods such as descriptive analytics, predictive analytics, and prescriptive analytics. Emphasis is placed on how these analytical techniques enhance the abstraction process, providing statistical rigor and predictive power to unveil hidden insights.The integration of data visualization and analytics is a focal point, showcasing their collective impact on various stages of data mining. From exploratory data analysis (EDA) for initial dataset understanding to the evaluation of mining models, the survey illuminates the collaborative nature of these components. Interactive dashboards emerge as a powerful tool, allowing users to dynamically explore datasets, visualize trends, and perform real-time analytics. Key Words:Data visualization , Data analytics ,Data mining , Tools and Software , Metrics and Key Performance Indicators (KPIs).

PelaSIG, a QGIS plugin for marine megafauna census: application to the aerial ACCOBAMS Survey Initiative dataset

The distribution of wild animals and their monitoring over large areas raises many logistical and technical difficulties that hinder the collection of observation data. The use of Geographic Information Systems (GIS) has increased significantly in recent years. QGIS, an open-source GIS software dedicated to the processing of geospatial data, enables the development of dedicated plugins for specific workflows. The open-source PelaSIG plugin has been developed in Python for QGIS 3 to facilitate and standardise the different steps before and after distance sampling surveys. It brings together a set of tools for survey preparation, automatic data checking, visualisation and presentation of survey effort and sightings to provide an adapted workflow. This plugin is currently designed to process dedicated aerial datasets collected with the SAMMOA software during marine megafauna surveys (i.e., marine mammals, seabirds, elasmobranchs, sea turtles, etc.). Here, we first describe the different tools already available, and then, we present an application with the dataset from the aerial survey of the ACCOBAMS Survey Initiative (ASI) conducted in 2018 over the Mediterranean Sea and using a multi-target protocol.

Modeling genotype-protein interaction and correlation for Alzheimer's disease: a multi-omics imaging genetics study.

Integrating and analyzing multiple omics data sets, including genomics, proteomics and radiomics, can significantly advance researchers' comprehensive understanding of Alzheimer's disease (AD). However, current methodologies primarily focus on the main effects of genetic variation and protein, overlooking non-additive effects such as genotype-protein interaction (GPI) and correlation patterns in brain imaging genetics studies. Importantly, these non-additive effects could contribute to intermediate imaging phenotypes, finally leading to disease occurrence. In general, the interaction between genetic variations and proteins, and their correlations are two distinct biological effects, and thus disentangling the two effects for heritable imaging phenotypes is of great interest and need. Unfortunately, this issue has been largely unexploited. In this paper, to fill this gap, we propose $\textbf{M}$ulti-$\textbf{T}$ask $\textbf{G}$enotype-$\textbf{P}$rotein $\textbf{I}$nteraction and $\textbf{C}$orrelation disentangling method ($\textbf{MT-GPIC}$) to identify GPI and extract correlation patterns between them. To ensure stability and interpretability, we use novel and off-the-shelf penalties to identify meaningful genetic risk factors, as well as exploit the interconnectedness of different brain regions. Additionally, since computing GPI poses a high computational burden, we develop a fast optimization strategy for solving MT-GPIC, which is guaranteed to converge. Experimental results on the Alzheimer's Disease Neuroimaging Initiative data set show that MT-GPIC achieves higher correlation coefficients and classification accuracy than state-of-the-art methods. Moreover, our approach could effectively identify interpretable phenotype-related GPI and correlation patterns in high-dimensional omics data sets. These findings not only enhance the diagnostic accuracy but also contribute valuable insights into the underlying pathogenic mechanisms of AD.

ESTSS—energy system time series suite: a declustered, application-independent, semi-artificial load profile benchmark set

This paper introduces an univariate application-independent set of load profiles or time series derived from real-world energy system data. The generation involved a two-step process: manifolding the initial dataset through signal processors to increase diversity and heterogeneity, followed by a declustering process that removes data redundancy. The study employed common feature engineering and machine learning techniques: the time series are transformed into a normalized feature space, followed by a dimensionality reduction via hierarchical clustering, and optimization. The resulting dataset is uniformly distributed across multiple feature space dimensions while retaining typical time and frequency domain characteristics inherent in energy system time series. This data serves various purposes, including algorithm testing, uncovering functional relationships between time series features and system performance, and training machine learning models. Two case studies demonstrate the claims: one focused on the suitability of hybrid energy storage systems and the other on quantifying the onsite hydrogen supply cost in green hydrogen production sites. The declustering algorithm, although a bys study, shows promise for further scientific exploration. The data and source code are openly accessible, providing a robust platform for future comparative studies. This work also offers smaller subsets for computationally intensive research. Data and source code can be found at https://github.com/s-guenther/estss and https://zenodo.org/records/10213145.

An AI-based framework for earthquake relief demand forecasting: A case study in Türkiye

Accurately forecasting the demand for relief assistance is crucial for efficient post-disaster relief operations. This paper presents an AI-based framework for earthquake relief demand forecasting, aiming to optimize the distribution of emergency resources. To identify higher-risk areas, the framework integrates historical earthquake data with socioeconomic information. This information aids in identifying regions that require immediate relief assistance. The initial dataset for relief demand quantification includes essential parameters such as Latitude, Longitude, Population, Demand, Supply, Cost, Deaths, Injuries, Infrastructural Damage, and Cascading Disaster. By analyzing these inputs and factors, AI algorithms can quantify the relief demand scenario accurately, resulting in accurate and reliable forecasts of relief supply requirements. The iterative ML techniques in generating comprehensive data for Türkiye greatly enhance the quality and accuracy of relief demand prediction. The prediction models used in the framework, namely Least Angle Regression and Linear Regression, demonstrate exceptional performance in forecasting relief demand. They achieve a remarkable R2 score of 1, indicating a perfect fit to the data, and a Loss Error function value of zero, signifying highly accurate predictions. Additionally, the Decision Tree algorithm showcases strong capabilities in iterative learning, achieving an accuracy rate of 88.99 %. In addition to providing valuable insights for tailored relief strategies, the framework extends beyond relief demand by predicting outcomes such as casualties, injuries, and infrastructure damage. With a comprehensive understanding of the earthquake’s impact, relief organizations can strategize their response and allocate resources accordingly. The thorough methodology and innovative application of ML algorithms within our pioneering AI-based framework significantly contribute to the field. This signifies a fundamental shift in the approach to envisioning and executing relief operations following seismic events.

Multimodal diagnosis model of Alzheimer’s disease based on improved Transformer

PurposeRecent technological advancements in data acquisition tools allowed neuroscientists to acquire different modality data to diagnosis Alzheimer’s disease (AD). However, how to fuse these enormous amount different modality data to improve recognizing rate and find significance brain regions is still challenging.MethodsThe algorithm used multimodal medical images [structural magnetic resonance imaging (sMRI) and positron emission tomography (PET)] as experimental data. Deep feature representations of sMRI and PET images are extracted by 3D convolution neural network (3DCNN). An improved Transformer is then used to progressively learn global correlation information among features. Finally, the information from different modalities is fused for identification. A model-based visualization method is used to explain the decisions of the model and identify brain regions related to AD.ResultsThe model attained a noteworthy classification accuracy of 98.1% for Alzheimer’s disease (AD) using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) dataset. Upon examining the visualization results, distinct brain regions associated with AD diagnosis were observed across different image modalities. Notably, the left parahippocampal region emerged consistently as a prominent and significant brain area.ConclusionsA large number of comparative experiments have been carried out for the model, and the experimental results verify the reliability of the model. In addition, the model adopts a visualization analysis method based on the characteristics of the model, which improves the interpretability of the model. Some disease-related brain regions were found in the visualization results, which provides reliable information for AD clinical research.

Integrating Ensemble Weather Predictions in a Hydrologic-Hydraulic Modelling System for Fine-Resolution Flood Forecasting: The Case of Skala Bridge at Evrotas River, Greece

Ensemble weather forecasting involves the integration of multiple simulations to improve the accuracy of predictions by introducing a probabilistic approach. It is difficult to accurately predict heavy rainfall events that cause flash floods and, thus, ensemble forecasting could be useful to reduce uncertainty in the forecast, thus improving emergency response. In this framework, this study presents the efforts to develop and assess a flash flood forecasting system that combines meteorological, hydrological, and hydraulic modeling, adopting an ensemble approach. The integration of ensemble weather forecasting and, subsequently, ensemble hydrological-hydraulic modeling can improve the accuracy of flash flood predictions, providing useful probabilistic information. The flash flood that occurred on 26 January 2023 in the Evrotas river basin (Greece) is used as a case study. The meteorological model, using 33 different initial and boundary condition datasets, simulated heavy rainfall, the hydrological model, using weather inputs, simulated discharge, and the hydraulic model, using discharge data, estimated water level at a bridge. The results show that the ensemble modeling system results in timely forecasts, while also providing valuable flooding probability information for 1 to 5 days prior, thus facilitating bridge flood warning. The continued refinement of such ensemble multi-model systems will further enhance the effectiveness of flash flood predictions and ultimately save lives and property.

Machine learning assisted discovery of high-efficiency self-healing epoxy coating for corrosion protection

Machine learning is a powerful means for the rapid development of high-performance functional materials. In this study, we presented a machine learning workflow for predicting the corrosion resistance of a self-healing epoxy coating containing ZIF-8@Ca microfillers. The orthogonal Latin square method was used to investigate the effects of the molecular weight of the polyetheramine curing agent, molar ratio of polyetheramine to epoxy, molar content of the hydrogen bond unit (UPy-D400), and mass content of the solid microfillers (ZIF-8@Ca microfillers) on the low impedance modulus (lg|Z|0.01Hz) values of the scratched coatings, generating 32 initial datasets. The machine learning workflow was divided into two stages: In stage I, five models were compared and the random forest (RF) model was selected for the active learning. After 5 cycles of active learning, the RF model achieved good prediction accuracy: coefficient of determination (R2) = 0.709, mean absolute percentage error (MAPE) = 0.081, root mean square error (RMSE) = 0.685 (lg(Ω·cm2)). In stage II, the best coating formulation was identified by Bayesian optimization. Finally, the electrochemical impedance spectroscopy (EIS) results showed that compared with the intact coating ((4.63 ± 2.08) × 1011 Ω·cm2), the |Z|0.01Hz value of the repaired coating was as high as (4.40 ± 2.04) × 1011 Ω·cm2. Besides, the repaired coating showed minimal corrosion and 3.3% of adhesion loss after 60 days of neutral salt spray testing.

Machine Learning for Predicting Neurodevelopmental Disorders in Children

Developmental domains like physical, verbal, cognitive, and social-emotional skills are crucial for monitoring a child’s growth. However, identifying neurodevelopmental deficiencies can be challenging due to the high level of variability and overlap. Early detection is essential, and digital procedures can assist in the process. This study leverages the current advances in artificial intelligence to address the prediction of neurodevelopmental disorders through a comprehensive machine learning approach. A novel and recently developed serious game dataset, collecting various data on children’s speech and linguistic responses, was used. The initial dataset comprised 520 instances, reduced to 473 participants after rigorous data preprocessing. Cluster analysis revealed distinct patterns and structures in the data, while reliability analysis ensured measurement consistency. A robust prediction model was developed using logistic regression. Applied to a subset of 184 participants with an average age of 7 years, the model demonstrated high accuracy, precision, recall, and F1-score, effectively distinguishing between instances with and without neurodevelopmental disorders. In conclusion, this research highlights the effectiveness of the machine learning approach in diagnosing neurodevelopmental disorders based on cognitive features, and offers new opportunities for decision making, classification, and clinical assessment, paving the way for early and personalized interventions for at-risk individuals.

Ionic liquid-assisted preparation of Ag–Zn–In–S quaternary quantum dot thin films and luminescence performance optimized by machine learning

Machine learning-assisted synthesis of new materials has been an ongoing hot topic. Luminescent Ag–Zn–In–S quaternary semiconductor quantum dot thin films have great application potential in many fields because of their excellent photoelectric performance. In this study, an environmentally friendly ionic liquid-assisted approach was employed to prepare high-performance Ag–Zn–In–S luminescent thin films, and machine learning methods were applied to screen the optimal preparation conditions. The photoluminescence (PL) performance, crystal structure, and morphology of the thin films were also been studied. The initial data set was obtained by high-throughput experiments. By comparing machine learning algorithms of Xgboost, Random Forest, Bagging, SVM, Extra Tree, and Gradient Boosting, the Random Forest achieved the best accuracy (r = 0.84, R2 = 0.847) and was selected to predict the optimal preparation conditions of Ag–Zn–In–S quaternary quantum dot thin films. The machine learning predicted results are in great agreement with the subsequent experimental results, which verifies the accuracy and effectiveness of the machine learning model. This study offers a theoretical foundation for preparing optimal inorganic quantum dot luminescent materials using machine learning.

Robust explanation supervision for false positive reduction in pulmonary nodule detection.

Lung cancer is the deadliest and second most common cancer in the United States due to the lack of symptoms for early diagnosis. Pulmonary nodules are small abnormal regions that can be potentially correlated to the occurrence of lung cancer. Early detection of these nodules is critical because it can significantly improve the patient's survival rates. Thoracic thin-sliced computed tomography (CT) scanning has emerged as a widely used method for diagnosing and prognosis lung abnormalities. The standard clinical workflow of detecting pulmonary nodules relies on radiologists to analyze CT images to assess the risk factors of cancerous nodules. However, this approach can be error-prone due to the various nodule formation causes, such as pollutants and infections. Deep learning (DL) algorithms have recently demonstrated remarkable success in medical image classification and segmentation. As an ever more important assistant to radiologists in nodule detection, it is imperative ensure the DL algorithm and radiologist to better understand the decisions from each other. This study aims to develop a framework integrating explainable AI methods to achieve accurate pulmonary nodule detection. A robust and explainable detection (RXD) framework is proposed, focusing on reducing false positives in pulmonary nodule detection. Its implementation is based on an explanation supervision method, which uses nodule contours of radiologists as supervision signals to force the model to learn nodule morphologies, enabling improved learning ability on small dataset, and enable small dataset learning ability. In addition, two imputation methods are applied to the nodule region annotations to reduce the noise within human annotations and allow the model to have robust attributions that meet human expectations. The 480, 265, and 265 CT image sets from the public Lung Image Database Consortium and Image Database Resource Initiative (LIDC-IDRI) dataset are used for training, validation, and testing. Using only 10, 30, 50, and 100 training samples sequentially, our method constantly improves the classification performance and explanation quality of baseline in terms of Area Under the Curve (AUC) and Intersection over Union (IoU). In particular, our framework with a learnable imputation kernel improves IoU from baseline by 24.0% to 80.0%. A pre-defined Gaussian imputation kernel achieves an even greater improvement, from 38.4% to 118.8% from baseline. Compared to the baseline trained on 100 samples, our method shows less drop in AUC when trained on fewer samples. A comprehensive comparison of interpretability shows that our method aligns better with expert opinions. A pulmonary nodule detection framework was demonstrated using public thoracic CT image datasets. The framework integrates the robust explanation supervision (RES) technique to ensure the performance of nodule classification and morphology. The method can reduce the workload of radiologists and enable them to focus on the diagnosis and prognosis of the potential cancerous pulmonary nodules at the early stage to improve the outcomes for lung cancer patients.

Pengelolaan Daun Kering untuk Pupuk Kompos

A material given to plants to increase nutrient availability is called fertilizer. This research project aims to find a solution to the environmental pollution problem caused by the accumulation of dry leaf litter. Compost is made using dry leaf litter waste as the main component and EM4 as an aid. EM4 is responsible for breaking down the organic material in the leaves. The finished dry leaf compost is odorless, in the form of loose granules, has a brittle and somewhat rough texture, and is dark brown in color. The research method used is literature research, or research derived from books and print journals accessible on Google Scholar. This type of research uses descriptive studies and several works of researchers with relevant fields of study. Data collected through inductive methods are then analyzed using one or more initial data sets. The obtained data is examined using a descriptive analytical approach. By transforming dry leaves into more useful and economically valuable compost, this research also aims to develop and enhance the creative abilities of the community and the local economy. Bright colors indicate composting that is too dry or moisture levels below 30%. One determining factor of compost maturity is its aroma. The compost fermentation process will produce various strong aromas.

Machine learning algorithms-based decision support model for diabetes

This research explores the application of machine learning (ML)-based risk prediction models in early diabetes disease detection for healthcare professionals. Diabetes affects millions of people worldwide. In light of significant advancements in biomedical sciences, vast volumes of data have been generated, including high-throughput genetic and diagnostic data sourced from extensive health records. Leveraging an initial diabetes risk prediction dataset from the University of California Irvine (UCI) ML repository, our research focused on supervised learning techniques, constituting 85% of the employed methods. The remaining 15% comprised unsupervised learning approaches, specifically association rules. A key contribution of this study lies in the development of an optimal prediction model utilizing supervised ML algorithms. The Boruta feature selection algorithm was employed to identify pertinent features, and the subsequent models were validated using a preprocessed dataset containing 10 attributes. Notably, the risk prediction models generated through random forest, extreme gradient boosting (XGBoost), and light gradient boosting machine (LightGBM) exhibited impressive average accuracies of 98.13%, 97.37%, and 97.22%, respectively, as determined via 10-fold cross-validation with 15 repetitions. Furthermore, these models achieved exceptional area under the ROC curve (AUC) values of 1, 0.99, and 0.99, respectively, showcasing their robustness and efficacy in diabetes risk prediction.

Anomaly detection in feature space for detecting changes in phytoplankton populations

Plankton organisms are fundamental components of the earth’s ecosystem. Zooplankton feeds on phytoplankton and is predated by fish and other aquatic animals, being at the core of the aquatic food chain. On the other hand, Phytoplankton has a crucial role in climate regulation, has produced almost 50% of the total oxygen in the atmosphere and it’s responsible for fixing around a quarter of the total earth’s carbon dioxide. Importantly, plankton can be regarded as a good indicator of environmental perturbations, as it can react to even slight environmental changes with corresponding modifications in morphology and behavior. At a population level, the biodiversity and the concentration of individuals of specific species may shift dramatically due to environmental changes. Thus, in this paper, we propose an anomaly detection-based framework to recognize heavy morphological changes in phytoplankton at a population level, starting from images acquired in situ. Given that an initial annotated dataset is available, we propose to build a parallel architecture training one anomaly detection algorithm for each available class on top of deep features extracted by a pre-trained Vision Transformer, further reduced in dimensionality with PCA. We later define global anomalies, corresponding to samples rejected by all the trained detectors, proposing to empirically identify a threshold based on global anomaly count over time as an indicator that can be used by field experts and institutions to investigate potential environmental perturbations. We use two publicly available datasets (WHOI22 and WHOI40) of grayscale microscopic images of phytoplankton collected with the Imaging FlowCytobot acquisition system to test the proposed approach, obtaining high performances in detecting both in-class and out-of-class samples. Finally, we build a dataset of 15 classes acquired by the WHOI across four years, showing that the proposed approach’s ability to identify anomalies is preserved when tested on images of the same classes acquired across a timespan of years.

An optimization framework for wind farm layout design using CFD-based Kriging model

Wind farm layout optimization (WFLO) seeks to alleviate the wake loss and maximize wind farm power output efficiency, and is a crucial process in the design and planning of wind energy projects. Since the optimization algorithms typically require thousands of numerical evaluations of the wake effects, conventional WFLO studies are usually carried out with the low-fidelity analytical wake models, while the higher-fidelity computational-fluid-dynamics-based (CFD-based) methods are seldom used due to the excessive computational cost. In this paper, we develop a self-adaptive optimization framework for wind farm layout design using CFD-based Kriging model to maximize the annual energy production (AEP) of wind farms. This surrogate-based optimization (SBO) framework uses latin hypercube sampling to generate a group of wind farm layout samples, based on which CFD simulations with the turbines modeled as actuator disks are carried out to obtain the corresponding AEPs. This initial wind farm layout dataset is used to train the Kriging model, which is then integrated with an optimizer based on genetic algorithm (GA). As the optimization progresses, the intermediate optimal layout designs are again fed into the dataset to re-train the Kriging model. Such adaptive update of wind farm layout dataset continues until the algorithm converges to the optimal layout design. To evaluate the performance of the proposed SBO framework, we apply it to four representative wind farm cases. Compared to the conventional staggered layout along the dominant wind direction, the optimized wind farm produces significantly higher total AEP, which is more evenly distributed among the turbines. In particular, the SBO framework requires significantly smaller number of CFD calls to yield the optimal layouts that generates almost the same AEP with the direct CFD-GA method. Further analysis on the velocity fields show that the optimization framework always attempts to locate the downstream turbines away from the wakes of upstream ones along the dominant wind directions. The proposed CFD-based surrogate model provides a more accurate and flexible alternative to the conventional analytical-wake-model-based methods in WFLO tasks, and has the potential to be used for designing efficient wind farm projects.

Hot Strip Mill Gearbox Monitoring and Diagnosis Based on Convolutional Neural Networks Using the Pseudo-Labeling Method

The steel industry is typical process manufacturing, and the quality and cost of the products can be improved by efficient operation of equipment. This paper proposes an efficient diagnosis and monitoring method for the gearbox, which is a key piece of mechanical equipment in steel manufacturing. In particular, an equipment maintenance plan for stable operation is essential. Therefore, equipment monitoring and diagnosis to prevent unplanned plant shutdowns are important to operate the equipment efficiently and economically. Most plant data collected on-site have no precise information about equipment malfunctions. Therefore, it is difficult to directly apply supervised learning algorithms to diagnose and monitor the equipment with the operational data collected. The purpose of this paper is to propose a pseudo-label method to enable supervised learning for equipment data without labels. Pseudo-normal (PN) and pseudo-abnormal (PA) vibration datasets are defined and labeled to apply classification analysis algorithms to unlabeled equipment data. To find an anomalous state in the equipment based on vibration data, the initial PN vibration dataset is compared with a PA vibration dataset collected over time, and the equipment is monitored for potential failure. Continuous wavelet transform (CWT) is applied to the vibration signals collected to obtain an image dataset, which is then entered into a convolutional neural network (an image classifier) to determine classification accuracy and detect equipment abnormalities. As a result of Steps 1 to 4, abnormal signals have already been detected in the dataset, and alarms and warnings have already been generated. The classification accuracy was over 0.95 at d=4, confirming quantitatively that the status of the equipment had changed significantly. In this way, a catastrophic failure can be avoided by performing a detailed equipment inspection in advance. Lastly, a catastrophic failure occurred in Step 9, and the classification accuracy ranged from 0.95 to 1.0. It was possible to prevent secondary equipment damage, such as motors connected to gearboxes, by identifying catastrophic failures promptly. This case study shows that the proposed procedure gives good results in detecting operation abnormalities of key unit equipment. In the conclusion, further promising topics are discussed.