During the storehouse surface rolling construction of a core rockfilldam, the spreading thickness of dam face is an important factor that affects the construction quality of the dam storehouse' rolling surface and...During the storehouse surface rolling construction of a core rockfilldam, the spreading thickness of dam face is an important factor that affects the construction quality of the dam storehouse' rolling surface and the overallquality of the entire dam. Currently, the method used to monitor and controlspreading thickness during the dam construction process is artificialsampling check after spreading, which makes it difficult to monitor the entire dam storehouse surface. In this paper, we present an in-depth study based on real-time monitoring and controltheory of storehouse surface rolling construction and obtain the rolling compaction thickness by analyzing the construction track of the rolling machine. Comparatively, the traditionalmethod can only analyze the rolling thickness of the dam storehouse surface after it has been compacted and cannot determine the thickness of the dam storehouse surface in realtime. To solve these problems, our system monitors the construction progress of the leveling machine and employs a real-time spreading thickness monitoring modelbased on the K-nearest neighbor algorithm. Taking the LHK core rockfilldam in Southwest China as an example, we performed real-time monitoring for the spreading thickness and conducted real-time interactive queries regarding the spreading thickness. This approach provides a new method for controlling the spreading thickness of the core rockfilldam storehouse surface.展开更多
The core of smoothed particle hydrodynamics （SPH） is the nearest neighbor search subroutine. In this paper, a nearest neighbor search algorithm which is based on multiple background grids and support variable smooth...The core of smoothed particle hydrodynamics （SPH） is the nearest neighbor search subroutine. In this paper, a nearest neighbor search algorithm which is based on multiple background grids and support variable smooth length is introduced. Through tested on lid driven cavity flow, it is clear that this method can provide high accuracy. Analysis and experiments have been made on its parallelism, and the results show that this method has better parallelism and with adding processors its accuracy become higher, thus it achieves that efficiency grows in pace with accuracy.展开更多
In this article,a new optimization system that uses few features to recognize locomotion with high classification accuracy is proposed.The optimization system consists of three parts.First,the features of the mixed me...In this article,a new optimization system that uses few features to recognize locomotion with high classification accuracy is proposed.The optimization system consists of three parts.First,the features of the mixed mechanical signal data are extracted from each analysis window of 200 ms after each foot contact event.Then,the Binary version of the hybrid Gray Wolf Optimization and Particle Swarm Optimization(BGWOPSO)algorithm is used to select features.And,the selected features are optimized and assigned different weights by the Biogeography-Based Optimization(BBO)algorithm.Finally,an improved K-Nearest Neighbor(KNN)classifier is employed for intention recognition.This classifier has the advantages of high accuracy,few parameters as well as low memory burden.Based on data from eight patients with transfemoral amputations,the optimization system is evaluated.The numerical results indicate that the proposed model can recognize nine daily locomotion modes(i.e.,low-,mid-,and fast-speed level-ground walking,ramp ascent/decent,stair ascent/descent,and sit/stand)by only seven features,with an accuracy of 96.66%±0.68%.As for real-time prediction on a powered knee prosthesis,the shortest prediction time is only 9.8 ms.These promising results reveal the potential of intention recognition based on the proposed system for high-level control of the prosthetic knee.展开更多
The Bugaji area,situated within the Malumfashi Schist Belt of northwestern Nigeria,primarily consists of metasediments that include quartzo-feldspathic and pelitic schists,and gneiss.However,this area poses a challeng...The Bugaji area,situated within the Malumfashi Schist Belt of northwestern Nigeria,primarily consists of metasediments that include quartzo-feldspathic and pelitic schists,and gneiss.However,this area poses a challenge in mineral exploration due to limited outcrop exposures and complex subsurface structures.Hence,there is the need for exhaustive geophysical studies and supplementary approaches to accurately delineate lithologies and structures.Therefore,this study combines field mapping and geophysical techniques with artificial intelligence(AI)modeling,comprising supervised learning algorithms,to overcome this exploration problem.Utilizing sophisticated AI techniques,specifically the Random Forest Classifier and K-Nearest Neighbor algorithms,geophysical data(gravity,magnetic,and radiometric measurements)were processed and analyzed.The AI model effectively filled data gaps,and identified potential lithological variations and prospective mineralization zones based on geophysical signatures derived from the integrated dataset.The AI modeling's commendable average accuracy of 85%in predicting values underscores its efficacy in interpreting geophysical data.The success of random forest in the geological mapping process can be attributed to its ability to handle high-dimensional data,capture non-linear relationships between input variables,and mitigate overfitting.The integrated approach enhanced our understanding of subsurface geology in the Bugaji area.展开更多
Variational mode decomposition(VMD) has been proved to be useful for extraction of fault-induced transients of rolling bearings. Multi-bandwidth mode manifold(Triple M, TM) is one variation of the VMD, which units mul...Variational mode decomposition(VMD) has been proved to be useful for extraction of fault-induced transients of rolling bearings. Multi-bandwidth mode manifold(Triple M, TM) is one variation of the VMD, which units multiple fault-related modes with different bandwidths by a nonlinear manifold learning algorithm named local tangent space alignment(LTSA). The merit of the TM method is that the bearing fault-induced transients extracted contain low level of in-band noise without optimization of the VMD parameters. However, the determination of the neighborhood size of the LTSA is time-consuming, and the extracted fault-induced transients may have the problem of asymmetry in the up-and-down direction. This paper aims to improve the efficiency and waveform symmetry of the TM method.Specifically, the multi-bandwidth modes consisting of the fault-related modes with different bandwidths are first obtained by repeating the recycling VMD(RVMD) method with different bandwidth balance parameters. Then, the LTSA algorithm is performed on the multi-bandwidth modes to extract their inherent manifold structure, in which the natural nearest neighbor(Triple N, TN) algorithm is adopted to efficiently and reasonably select the neighbors of each data point in the multi-bandwidth modes. Finally, a weight-based feature compensation strategy is designed to synthesize the low-dimensional manifold features to alleviate the asymmetry problem, resulting in a symmetric TM feature that can represent the real fault transient components. The major contribution of the improved TM method for bearing fault diagnosis is that the pure fault-induced transients are extracted efficiently and are symmetrical as the real. One simulation analysis and two experimental applications in bearing fault diagnosis validate the enhanced performance of the improved TM method over the traditional methods. This research proposes a bearing fault diagnosis method which has the advantages of high efficiency, good waveform symmetry and enhanced in-band noise removal capability.展开更多
To address cross-ISP traffic problem caused by BitTorrent,we present our design and evaluation of a proximity-aware BitTorrent system. In our approach,clients generate global proximity-aware information by using landm...To address cross-ISP traffic problem caused by BitTorrent,we present our design and evaluation of a proximity-aware BitTorrent system. In our approach,clients generate global proximity-aware information by using landmark clustering;the tracker uses this proximity to maintain all peers in an orderly way and hands back a biased subset consisting of the peers who are physically closest to the requestor. Our approach requires no co-operation between P2P users and their Internet infra structures,such as ISPs or CDNs,no constantly path monitoring or probing their neighbors. The simulation results show that our approach can not only reduce unnecessary cross-ISP traffic,but also allow downloadsing fast.展开更多
Slurry electrolysis(SE),as a hydrometallurgical process,has the characteristic of a multitank series connection,which leads to various stirring conditions and a complex solid suspension state.The computational fluid d...Slurry electrolysis(SE),as a hydrometallurgical process,has the characteristic of a multitank series connection,which leads to various stirring conditions and a complex solid suspension state.The computational fluid dynamics(CFD),which requires high computing resources,and a combination with machine learning was proposed to construct a rapid prediction model for the liquid flow and solid concentration fields in a SE tank.Through scientific selection of calculation samples via orthogonal experiments,a comprehensive dataset covering a wide range of conditions was established while effectively reducing the number of simulations and providing reasonable weights for each factor.Then,a prediction model of the SE tank was constructed using the K-nearest neighbor algorithm.The results show that with the increase in levels of orthogonal experiments,the prediction accuracy of the model improved remarkably.The model established with four factors and nine levels can accurately predict the flow and concentration fields,and the regression coefficients of average velocity and solid concentration were 0.926 and 0.937,respectively.Compared with traditional CFD,the response time of field information prediction in this model was reduced from 75 h to 20 s,which solves the problem of serious lag in CFD applied alone to actual production and meets real-time production control requirements.展开更多
The identification of targets varies in different surge tests. A multi-color space threshold segmentation and self-learning k-nearest neighbor algorithm （k-NN） for equipment under test status identification was prop...The identification of targets varies in different surge tests. A multi-color space threshold segmentation and self-learning k-nearest neighbor algorithm （k-NN） for equipment under test status identification was proposed after using feature matching to identify equipment status had to train new patterns every time before testing. First, color space （L＊a＊b＊, hue saturation lightness （HSL）, hue saturation value （HSV）） to segment was selected according to the high luminance points ratio and white luminance points ratio of the image. Second, the unknown class sample Sr was classified by the k-NN algorithm with training set T~ according to the feature vector, which was formed from number ofpixels, eccentricity ratio, compact- ness ratio, and Euler＇s numbers. Last, while the classification confidence coefficient equaled k, made Sr as one sample ofpre-training set Tz＇. The training set Tz increased to Tz＋1 by Tz＇ if Tz＇ was saturated. In nine series of illuminant, indicator light, screen, and disturbances samples （a total of 21600 frames）, the algorithm had a 98.65% identification accuracy, also selected five groups of samples to enlarge the training set from To to T5 by itself. Keywords multi-color space, k-nearest neighbor algorithm （k-NN）, self-learning, surge test展开更多
The complexity and unpredictability of clear air turbulence(CAT)pose significant challenges to aviation safety.Accurate prediction of turbulence events is crucial for reducing flight accidents and economic losses.Howe...The complexity and unpredictability of clear air turbulence(CAT)pose significant challenges to aviation safety.Accurate prediction of turbulence events is crucial for reducing flight accidents and economic losses.However,traditional turbulence prediction methods,such as ensemble forecasting techniques,have certain limitations:they only consider turbulence data from the most recent period,making it difficult to capture the nonlinear relationships present in turbulence.This study proposes a turbulence forecasting model based on the K-nearest neighbor(KNN)algorithm,which uses a combination of eight CAT diagnostic features as the feature vector and introduces CAT diagnostic feature weights to improve prediction accuracy.The model calculates the results of seven years of CAT diagnostics from 125 to 500 hPa obtained from the ECMWF fifth-generation reanalysis dataset(ERA5)as feature vector inputs and combines them with the labels of Pilot Reports(PIREP)annotated data,where each sample contributes to the prediction result.By measuring the distance between the current CAT diagnostic variable and other variables,the model determines the climatically most similar neighbors and identifies the turbulence intensity category caused by the current variable.To evaluate the model’s performance in diagnosing high-altitude turbulence over Colorado,PIREP cases were randomly selected for analysis.The results show that the weighted KNN(W-KNN)model exhibits higher skill in turbulence prediction,and outperforms traditional prediction methods and other machine learning models(e.g.,Random Forest)in capturing moderate or greater(MOG)level turbulence.The performance of the model was confirmed by evaluating the receiver operating characteristic(ROC)curve,maximum True Skill Statistic(maxTSS=0.552),and reliability plot.A robust score(area under the curve:AUC=0.86)was obtained,and the model demonstrated sensitivity to seasonal and annual climate fluctuations.展开更多
Accurate traffic state estimations(TSEs)within road networks are crucial for enhancing intelligent transportation systems and developing effective traffic management strategies.Traditional TSE methods often assume hom...Accurate traffic state estimations(TSEs)within road networks are crucial for enhancing intelligent transportation systems and developing effective traffic management strategies.Traditional TSE methods often assume homogeneous traffic,where all vehicles are considered identical,which does not accurately reflect the complexities of real traffic conditions that often exhibit heterogeneous characteristics.In this study,we address the limitations of conventional models by introducing a novel TSE model designed for precise estimations of heterogeneous traffic flows.We develop a comprehensive traffic feature index system tailored for heterogeneous traffic that includes four elements:basic traffic parameters,heterogeneous vehicle speeds,heterogeneous vehicle flows,and mixed flow rates.This system aids in capturing the unique traffic characteristics of different vehicle types.Our proposed high-dimensional fuzzy TSE model,termed HiF-TSE,integrates three main processes:feature selection,which eliminates redundant traffic features using Spearman correlation coefficients;dimension reduction,which utilizes the T-distributed stochastic neighbor embedding machine learning algorithm to reduce high-dimensional traffic feature data;and FCM clustering,which applies the fuzzy C-means algorithm to classify the simplified data into distinct clusters.The HiF-TSE model significantly reduces computational demands and enhances efficiency in TSE processing.We validate our model through a real-world case study,demonstrating its ability to adapt to variations in vehicle type compositions within heterogeneous traffic and accurately represent the actual traffic state.展开更多
In the process of obtaining information from the actual traffic network, the incomplete data set caused by missing data reduces the validity of the data and the performance of the data-driven model. A traffic flow rep...In the process of obtaining information from the actual traffic network, the incomplete data set caused by missing data reduces the validity of the data and the performance of the data-driven model. A traffic flow repair model based on a k-nearest neighbor(KNN) spatio-temporal attention(STA) graph convolutional network(KAGCN) was proposed in this paper. Firstly, the missing data is initially interpolated by the KNN algorithm, and then the complete index set(CIS) is constructed by combining the adjacency matrix of the network structure. Secondly, a STA mechanism is added to the CIS to capture the spatio-temporal correlation between the data. Then, the graph neural network(GNN) is used to reconstruct the data by spatio-temporal correlation, and the reconstructed data set is used to correct and optimize the initial interpolation data set to obtain the final repair result. Finally, the PEMSD4 data set is used to simulate the missing data in the actual road network, and experiments are carried out under the missing rate of 30%, 50%, and 70% respectively. The results show that the mean absolute error(MAE), root mean square error(RMSE), and mean absolute percentage error(MAPE) of the KAGCN model increased by at least 3.83%, 2.80%, and 5.33%, respectively, compared to the other baseline models at different deletion rates. It proves that the KAGCN model is effective in repairing the missing data of traffic flow.展开更多
This paper deals with two new methods,based on k-NN algorithm,for fault detection and classification in distance protection.In these methods,by finding the distance between each sample and its fifth nearest neighbor i...This paper deals with two new methods,based on k-NN algorithm,for fault detection and classification in distance protection.In these methods,by finding the distance between each sample and its fifth nearest neighbor in a predefault window,the fault occurrence time and the faulty phases are determined.The maximum value of the distances in case of detection and classification procedures is compared with pre-defined threshold values.The main advantages of these methods are:simplicity,low calculation burden,acceptable accuracy,and speed.The performance of the proposed scheme is tested on a typical system in MATLAB Simulink.Various possible fault types in different fault resistances,fault inception angles,fault locations,short circuit levels,X/R ratios,source load angles are simulated.In addition,the performance of similar six well-known classification techniques is compared with the proposed classification method using plenty of simulation data.展开更多
基金supported by the Innovative Research Groups of National Natural Science Foundation of China(No. 51621092)National Basic Research Program of China ("973" Program, No. 2013CB035904)National Natural Science Foundation of China (No. 51439005)
文摘During the storehouse surface rolling construction of a core rockfilldam, the spreading thickness of dam face is an important factor that affects the construction quality of the dam storehouse' rolling surface and the overallquality of the entire dam. Currently, the method used to monitor and controlspreading thickness during the dam construction process is artificialsampling check after spreading, which makes it difficult to monitor the entire dam storehouse surface. In this paper, we present an in-depth study based on real-time monitoring and controltheory of storehouse surface rolling construction and obtain the rolling compaction thickness by analyzing the construction track of the rolling machine. Comparatively, the traditionalmethod can only analyze the rolling thickness of the dam storehouse surface after it has been compacted and cannot determine the thickness of the dam storehouse surface in realtime. To solve these problems, our system monitors the construction progress of the leveling machine and employs a real-time spreading thickness monitoring modelbased on the K-nearest neighbor algorithm. Taking the LHK core rockfilldam in Southwest China as an example, we performed real-time monitoring for the spreading thickness and conducted real-time interactive queries regarding the spreading thickness. This approach provides a new method for controlling the spreading thickness of the core rockfilldam storehouse surface.
基金Project supported by the National Natural Science Foundation of China(Grant No.11002086)the Shanghai Leading Academic Discipline Project(Grant No.J50103)
文摘The core of smoothed particle hydrodynamics （SPH） is the nearest neighbor search subroutine. In this paper, a nearest neighbor search algorithm which is based on multiple background grids and support variable smooth length is introduced. Through tested on lid driven cavity flow, it is clear that this method can provide high accuracy. Analysis and experiments have been made on its parallelism, and the results show that this method has better parallelism and with adding processors its accuracy become higher, thus it achieves that efficiency grows in pace with accuracy.
基金This research was supported in part by the National Key Research and Development Program of China under Grant 2018YFC2001300in part by the National Natural Science Foundation of China under Grant 91948302,Grant 91848204,and Grant 52021003the Project of Scientific and Technological Development Plan of Jilin Province under Grant 20220508130RC.
文摘In this article,a new optimization system that uses few features to recognize locomotion with high classification accuracy is proposed.The optimization system consists of three parts.First,the features of the mixed mechanical signal data are extracted from each analysis window of 200 ms after each foot contact event.Then,the Binary version of the hybrid Gray Wolf Optimization and Particle Swarm Optimization(BGWOPSO)algorithm is used to select features.And,the selected features are optimized and assigned different weights by the Biogeography-Based Optimization(BBO)algorithm.Finally,an improved K-Nearest Neighbor(KNN)classifier is employed for intention recognition.This classifier has the advantages of high accuracy,few parameters as well as low memory burden.Based on data from eight patients with transfemoral amputations,the optimization system is evaluated.The numerical results indicate that the proposed model can recognize nine daily locomotion modes(i.e.,low-,mid-,and fast-speed level-ground walking,ramp ascent/decent,stair ascent/descent,and sit/stand)by only seven features,with an accuracy of 96.66%±0.68%.As for real-time prediction on a powered knee prosthesis,the shortest prediction time is only 9.8 ms.These promising results reveal the potential of intention recognition based on the proposed system for high-level control of the prosthetic knee.
文摘The Bugaji area,situated within the Malumfashi Schist Belt of northwestern Nigeria,primarily consists of metasediments that include quartzo-feldspathic and pelitic schists,and gneiss.However,this area poses a challenge in mineral exploration due to limited outcrop exposures and complex subsurface structures.Hence,there is the need for exhaustive geophysical studies and supplementary approaches to accurately delineate lithologies and structures.Therefore,this study combines field mapping and geophysical techniques with artificial intelligence(AI)modeling,comprising supervised learning algorithms,to overcome this exploration problem.Utilizing sophisticated AI techniques,specifically the Random Forest Classifier and K-Nearest Neighbor algorithms,geophysical data(gravity,magnetic,and radiometric measurements)were processed and analyzed.The AI model effectively filled data gaps,and identified potential lithological variations and prospective mineralization zones based on geophysical signatures derived from the integrated dataset.The AI modeling's commendable average accuracy of 85%in predicting values underscores its efficacy in interpreting geophysical data.The success of random forest in the geological mapping process can be attributed to its ability to handle high-dimensional data,capture non-linear relationships between input variables,and mitigate overfitting.The integrated approach enhanced our understanding of subsurface geology in the Bugaji area.
基金Supported by National Natural Science Foundation of China (Grant Nos. 51805342,51875376, 52007128)Jiangsu Provincial Natural Science Foundation of China (Grant No. BK20180842)+2 种基金China Postdoctoral Science Foundation (Grant Nos. 2021M692354, 2018M640514)Suzhou Prospective Research Program of China (Grant No. SYG201932)Jiangsu Provincial Natural Science Fund for Colleges and Universities of China (Grant No. 18KJB470022)。
文摘Variational mode decomposition(VMD) has been proved to be useful for extraction of fault-induced transients of rolling bearings. Multi-bandwidth mode manifold(Triple M, TM) is one variation of the VMD, which units multiple fault-related modes with different bandwidths by a nonlinear manifold learning algorithm named local tangent space alignment(LTSA). The merit of the TM method is that the bearing fault-induced transients extracted contain low level of in-band noise without optimization of the VMD parameters. However, the determination of the neighborhood size of the LTSA is time-consuming, and the extracted fault-induced transients may have the problem of asymmetry in the up-and-down direction. This paper aims to improve the efficiency and waveform symmetry of the TM method.Specifically, the multi-bandwidth modes consisting of the fault-related modes with different bandwidths are first obtained by repeating the recycling VMD(RVMD) method with different bandwidth balance parameters. Then, the LTSA algorithm is performed on the multi-bandwidth modes to extract their inherent manifold structure, in which the natural nearest neighbor(Triple N, TN) algorithm is adopted to efficiently and reasonably select the neighbors of each data point in the multi-bandwidth modes. Finally, a weight-based feature compensation strategy is designed to synthesize the low-dimensional manifold features to alleviate the asymmetry problem, resulting in a symmetric TM feature that can represent the real fault transient components. The major contribution of the improved TM method for bearing fault diagnosis is that the pure fault-induced transients are extracted efficiently and are symmetrical as the real. One simulation analysis and two experimental applications in bearing fault diagnosis validate the enhanced performance of the improved TM method over the traditional methods. This research proposes a bearing fault diagnosis method which has the advantages of high efficiency, good waveform symmetry and enhanced in-band noise removal capability.
基金supported in part by the National High-tech Research and Development Program (863 Program) of China under Grant No. 2009AA01Z210, No. 2009AA01Z250 and No. 2008AA01A324support from Guangdong Ministry of Education Industry-Academia-Research project No. 2009B090300315EU FP7 Project (INFSO-ICT- 215549)
文摘To address cross-ISP traffic problem caused by BitTorrent,we present our design and evaluation of a proximity-aware BitTorrent system. In our approach,clients generate global proximity-aware information by using landmark clustering;the tracker uses this proximity to maintain all peers in an orderly way and hands back a biased subset consisting of the peers who are physically closest to the requestor. Our approach requires no co-operation between P2P users and their Internet infra structures,such as ISPs or CDNs,no constantly path monitoring or probing their neighbors. The simulation results show that our approach can not only reduce unnecessary cross-ISP traffic,but also allow downloadsing fast.
基金financially supported by the National Natural Science Foundation of China(No.51974018the Open Foundation of the State Key Laboratory of Process Automation in Mining and Metallurgy(No.BGRIMM-KZSKL-2022-9).
文摘Slurry electrolysis(SE),as a hydrometallurgical process,has the characteristic of a multitank series connection,which leads to various stirring conditions and a complex solid suspension state.The computational fluid dynamics(CFD),which requires high computing resources,and a combination with machine learning was proposed to construct a rapid prediction model for the liquid flow and solid concentration fields in a SE tank.Through scientific selection of calculation samples via orthogonal experiments,a comprehensive dataset covering a wide range of conditions was established while effectively reducing the number of simulations and providing reasonable weights for each factor.Then,a prediction model of the SE tank was constructed using the K-nearest neighbor algorithm.The results show that with the increase in levels of orthogonal experiments,the prediction accuracy of the model improved remarkably.The model established with four factors and nine levels can accurately predict the flow and concentration fields,and the regression coefficients of average velocity and solid concentration were 0.926 and 0.937,respectively.Compared with traditional CFD,the response time of field information prediction in this model was reduced from 75 h to 20 s,which solves the problem of serious lag in CFD applied alone to actual production and meets real-time production control requirements.
文摘The identification of targets varies in different surge tests. A multi-color space threshold segmentation and self-learning k-nearest neighbor algorithm （k-NN） for equipment under test status identification was proposed after using feature matching to identify equipment status had to train new patterns every time before testing. First, color space （L＊a＊b＊, hue saturation lightness （HSL）, hue saturation value （HSV）） to segment was selected according to the high luminance points ratio and white luminance points ratio of the image. Second, the unknown class sample Sr was classified by the k-NN algorithm with training set T~ according to the feature vector, which was formed from number ofpixels, eccentricity ratio, compact- ness ratio, and Euler＇s numbers. Last, while the classification confidence coefficient equaled k, made Sr as one sample ofpre-training set Tz＇. The training set Tz increased to Tz＋1 by Tz＇ if Tz＇ was saturated. In nine series of illuminant, indicator light, screen, and disturbances samples （a total of 21600 frames）, the algorithm had a 98.65% identification accuracy, also selected five groups of samples to enlarge the training set from To to T5 by itself. Keywords multi-color space, k-nearest neighbor algorithm （k-NN）, self-learning, surge test
基金Supported by the Nanjing University of Aeronautics and Astronautics(KFB2305601).
文摘The complexity and unpredictability of clear air turbulence(CAT)pose significant challenges to aviation safety.Accurate prediction of turbulence events is crucial for reducing flight accidents and economic losses.However,traditional turbulence prediction methods,such as ensemble forecasting techniques,have certain limitations:they only consider turbulence data from the most recent period,making it difficult to capture the nonlinear relationships present in turbulence.This study proposes a turbulence forecasting model based on the K-nearest neighbor(KNN)algorithm,which uses a combination of eight CAT diagnostic features as the feature vector and introduces CAT diagnostic feature weights to improve prediction accuracy.The model calculates the results of seven years of CAT diagnostics from 125 to 500 hPa obtained from the ECMWF fifth-generation reanalysis dataset(ERA5)as feature vector inputs and combines them with the labels of Pilot Reports(PIREP)annotated data,where each sample contributes to the prediction result.By measuring the distance between the current CAT diagnostic variable and other variables,the model determines the climatically most similar neighbors and identifies the turbulence intensity category caused by the current variable.To evaluate the model’s performance in diagnosing high-altitude turbulence over Colorado,PIREP cases were randomly selected for analysis.The results show that the weighted KNN(W-KNN)model exhibits higher skill in turbulence prediction,and outperforms traditional prediction methods and other machine learning models(e.g.,Random Forest)in capturing moderate or greater(MOG)level turbulence.The performance of the model was confirmed by evaluating the receiver operating characteristic(ROC)curve,maximum True Skill Statistic(maxTSS=0.552),and reliability plot.A robust score(area under the curve:AUC=0.86)was obtained,and the model demonstrated sensitivity to seasonal and annual climate fluctuations.
基金supported by the National Key R&D Program of China(Grant No.2023YFB4302702)the Fundamental Research Funds for the Central Universities(Grant No.2023JKF02ZK08).
文摘Accurate traffic state estimations(TSEs)within road networks are crucial for enhancing intelligent transportation systems and developing effective traffic management strategies.Traditional TSE methods often assume homogeneous traffic,where all vehicles are considered identical,which does not accurately reflect the complexities of real traffic conditions that often exhibit heterogeneous characteristics.In this study,we address the limitations of conventional models by introducing a novel TSE model designed for precise estimations of heterogeneous traffic flows.We develop a comprehensive traffic feature index system tailored for heterogeneous traffic that includes four elements:basic traffic parameters,heterogeneous vehicle speeds,heterogeneous vehicle flows,and mixed flow rates.This system aids in capturing the unique traffic characteristics of different vehicle types.Our proposed high-dimensional fuzzy TSE model,termed HiF-TSE,integrates three main processes:feature selection,which eliminates redundant traffic features using Spearman correlation coefficients;dimension reduction,which utilizes the T-distributed stochastic neighbor embedding machine learning algorithm to reduce high-dimensional traffic feature data;and FCM clustering,which applies the fuzzy C-means algorithm to classify the simplified data into distinct clusters.The HiF-TSE model significantly reduces computational demands and enhances efficiency in TSE processing.We validate our model through a real-world case study,demonstrating its ability to adapt to variations in vehicle type compositions within heterogeneous traffic and accurately represent the actual traffic state.
基金supported by the National Natural Science Foundation of China (62162040)the Gansu Provincial Science and Technology Plan Funding Key Project of Natural Science Foundation of China (22JR5RA226)+1 种基金the Gansu Province Higher Education Innovation Fund-Funded Project (2021A-028)the Gansu Provincial Science and Technology Program Funding Project (21ZD4GA028)。
文摘In the process of obtaining information from the actual traffic network, the incomplete data set caused by missing data reduces the validity of the data and the performance of the data-driven model. A traffic flow repair model based on a k-nearest neighbor(KNN) spatio-temporal attention(STA) graph convolutional network(KAGCN) was proposed in this paper. Firstly, the missing data is initially interpolated by the KNN algorithm, and then the complete index set(CIS) is constructed by combining the adjacency matrix of the network structure. Secondly, a STA mechanism is added to the CIS to capture the spatio-temporal correlation between the data. Then, the graph neural network(GNN) is used to reconstruct the data by spatio-temporal correlation, and the reconstructed data set is used to correct and optimize the initial interpolation data set to obtain the final repair result. Finally, the PEMSD4 data set is used to simulate the missing data in the actual road network, and experiments are carried out under the missing rate of 30%, 50%, and 70% respectively. The results show that the mean absolute error(MAE), root mean square error(RMSE), and mean absolute percentage error(MAPE) of the KAGCN model increased by at least 3.83%, 2.80%, and 5.33%, respectively, compared to the other baseline models at different deletion rates. It proves that the KAGCN model is effective in repairing the missing data of traffic flow.
文摘This paper deals with two new methods,based on k-NN algorithm,for fault detection and classification in distance protection.In these methods,by finding the distance between each sample and its fifth nearest neighbor in a predefault window,the fault occurrence time and the faulty phases are determined.The maximum value of the distances in case of detection and classification procedures is compared with pre-defined threshold values.The main advantages of these methods are:simplicity,low calculation burden,acceptable accuracy,and speed.The performance of the proposed scheme is tested on a typical system in MATLAB Simulink.Various possible fault types in different fault resistances,fault inception angles,fault locations,short circuit levels,X/R ratios,source load angles are simulated.In addition,the performance of similar six well-known classification techniques is compared with the proposed classification method using plenty of simulation data.
