Flow Prediction Research Articles

Matrix-based long-term traffic flow prediction

ABSTRACT Accurate long-term traffic prediction is crucial for enhancing traffic efficiency, ensuring urban safety, and fostering sustainable urban development. However, forecasting over extended periods is challenging due to intricate trends, cyclical variations, and interference from outlier data. To address these issues, this study proposes a matrix-based traffic flow prediction model. The model constructs a matrix with periods as rows and similarities as columns, leveraging periodicity and similarity in traffic data. A row-column prediction module links these patterns efficiently, while a fluctuation transformation mitigates the impact of outliers, significantly improving forecast accuracy. Extending the forecast time span to 14 days with hourly intervals, the model was validated using the PeMS dataset provided by the California Department of Transportation. Results demonstrate the model’s effectiveness in capturing complex temporal dynamics, providing a robust tool for long-term traffic prediction.

Effects of Turbulence Modeling on the Simulation of Wind Flow over Typical Complex Terrains

The correct prediction of the wind speed and turbulence levels over complex terrain is essential for accurately assessing wind turbine wake recovery, power production, safety, and wind farm design. In this paper, two modified RANS turbulence models are proposed, which are innovative variants of the conventional SST k-ω model and the linear Reynolds stress model (RSM) featuring optimized closure constants. Then, these two modified models and their origin models are applied to compare and analyze wind flows from a 3D hill wind tunnel experiment and two field measurements over typical complex terrain, including Askervein hill and Bolund island, with the aim of analyzing the sensitivity of wind flows to different RANS turbulence models. The study focuses on analyzing the effects of different turbulence models on the self-sustainability of wind speed and turbulent kinetic energy upstream of the computational domain and on the accuracy of wind flow prediction over complex terrain. The results show that our modified RSM model shows better agreement with the available experimental data on the upstream and leeward sides of all simulated hills. The wind speed on the leeward slope is particularly sensitive to the turbulence model, with a maximum difference in the relative root mean square error (RRMSE) that can reach 11% among the four models. The accuracy of the turbulent kinetic energy depends on the self-sustainability of the upstream turbulent kinetic energy and the predictive ability of the turbulence model for separated flows, and the maximum difference in the RRMSE of the four models can reach 47%. In addition, the advantages and disadvantages of the tested models are discussed to provide guidance for model selection during wind flow simulations in complex terrain.

PREFER: A Pre-trained Model Recommendation Framework for Edge Computing Enabled Traffic Flow Prediction

The recent years have witnessed a surge in the development of traffic flow prediction methods, often deployed on cloud platforms to offer predictive services for entire transportation networks. However, the processes of training and executing a model for the entire traffic network are both time-consuming and computationally expensive. As a result, the utilization of edge servers for local sub-network prediction services has gained prominence. Nevertheless, training prediction models for numerous sub-networks within the extensive traffic network remains a time-intensive and computing resource-consuming task. To tackle this challenge, this paper introduces the Pre-trained model REcommendation Framework for traffic sub-nEtwork in tRaffic flow prediction (PREFER). PREFER trains a set of traffic flow prediction models on selected sub-networks, then recommends optimal pre-trained models for edge servers. The recommendation is specifically based on performance prediction, integrating neural collaborative filtering and traffic flow characteristics. Experiments conducted on real datasets reveal that the pre-trained models recommended by PREFER perform close to the actual optimal ones and significantly outperform existing recommendation algorithms.

Nano-fluid flow predictions in convergent/divergent channels using ANN-BLMT and physics-informed neural networks

Nano-fluid flow predictions in convergent/divergent channels using ANN-BLMT and physics-informed neural networks

Research on a Passenger Flow Prediction Model Based on BWO-TCLS-Self-Attention

In recent years, with the rapid development of the global demand and scale for deep underground space utilization, deep space has gradually transitioned from single-purpose uses such as underground transportation, civil defense, and commerce to a comprehensive, livable, and disaster-resistant underground ecosystem. This shift has brought increasing attention to the safety of personnel flow in deep spaces. In addressing challenges in deep space passenger flow prediction, such as irregular flow patterns, surges in extreme conditions, large data dimensions, and redundant features complicating the model, this paper proposes a deep space passenger flow prediction model that integrates a Temporal Convolutional Network (TCN) and Long Short-Term Memory (LSTM) network. The model first employs a dual-layer LSTM network structure with a Dropout layer to capture complex temporal dynamics while preventing overfitting. Then, a Self-Attention mechanism and TCN network are introduced to reduce redundant feature data and enhance the model’s performance and speed. Finally, the Beluga Whale Optimization (BWO) algorithm is used to optimize hyperparameters, further improving the prediction accuracy of the network. Experimental results demonstrate that the BWO-TCLS-Self-Attention model proposed in this paper achieves an R2 value of 96.94%, with MAE and RMSE values of 118.464 and 218.118, respectively. Compared with some mainstream prediction models, the R2 value has increased, while both MAE and RMSE values have decreased, indicating its ability to accurately predict passenger flow in deep underground spaces.

Short-Term Traffic Flow Prediction: Techniques and Methods

With the development of the economy, the expansion of the city scale, the increase in the population of urban residents and the increase in the per capita car ownership, the traffic congestion problem has an increasing impact on urban operation. Especially during the peak period of commuting traffic and holidays in large cities, most cities are facing the serious problem of road congestion. In order to facilitate the travel of residents and reduce air pollution, it is very important to keep the main road of the city smooth and rapid passage of vehicles. Therefore, cities should vigorously develop short-time traffic flow prediction, and accurate prediction of short-time traffic flow is the key to realizing intelligent urban traffic operation and efficient traffic flow management improvement. This paper will talk about the time series prediction model of short-time traffic flow prediction methods and advantages and disadvantages to provide auxiliary support for solving the traffic congestion problem, provide efficient travel path decision-making guidance for urban residents, facilitate the saving of urban people's travel costs, and also provide effective police deployment decision-making information for the traffic management department.

New Approaches of Traffic Flow Prediction

Recently, transportation has become a part of people's workday lives, with the fierce development of information, technology, and living standards. Nevertheless, it causes a great deal of inevitable trouble, such as congestion. As artificial intelligence (AI) is becoming increasingly popular, people have begun looking for solutions from AI daily. To bring convenience to relevant scientific research, the author has reviewed some novel AI models for traffic flow forecasting designed in recent years, based on a brief introduction to both machine learning and deep learning. According to the newly proposed AI models, the author has also drawn some expectations of traffic flow prediction that have the potential to be widely applied, the expectation combines mathematical functions, statistical models, and AI algorithms, making the traffic flow forecasting and adjustment more scientific and precise. The integration of advanced AI models into traffic flow prediction holds significant promise for enhancing urban mobility and reducing congestion in our daily lives.

Checkpoint data-driven GCN-GRU vehicle trajectory and traffic flow prediction

With the development of information technology, massive traffic data-driven short-term traffic situation analysis of urban road networks has become a research hotspot in urban traffic management. Accurate vehicle trajectory and traffic flow prediction can provide technical support for vehicle path planning and road congestion warning. Unlike most studies that use GPS data to predict vehicle trajectories, this paper combines the broad coverage, high reliability, and lighter weight of traffic checkpoint data to propose a method that uses trajectory prediction technology to forecast the traffic flow in urban road networks accurately. The method adopts a checkpoint data-driven approach for data collection, combines graph convolutional neural network (GCN) and gated recurrent unit (GRU) models to more effectively learn and extract spatiotemporal correlation features of vehicle trajectories, which significantly improves the accuracy of vehicle trajectory prediction, and uses the output of the trajectory prediction model to forecast traffic flow more accurately. Firstly, transforming the checkpoint data into daily vehicle trajectories with time series characteristics, realizing the vehicle trajectory travel chain division. Secondly, the adjacency matrix is established by using the spatial relationship of each checkpoint, and the feature matrix of the vehicle’s driving trajectory over time is established, which is used as the input of GCN to learn the spatial characteristics of the vehicle while driving on the road network, and then GRU is added to further process the data after GCN training, constructing a GCN-GRU vehicle trajectory prediction model for vehicle trajectory prediction. Finally, the traffic flow of each checkpoint is calculated based on the prediction result of vehicle trajectory and compared with the real checkpoint flow. This paper conducts many experiments on the Qingdao City Shinan district checkpoint dataset. The results show that compared with the single models GCN, GRU, BiGRU, and BiLSTM, the GCN-GRU model has reduced the MAE by 0.75, 0.46, 0.52, and 0.57, and the RMSE by 0.76, 0.52, 0.58, and 0.68, respectively, demonstrating stronger spatial and temporal correlation characteristics and higher prediction accuracy. The MAPE between the forecasted flow and the real flow is 0.18, which verifies the reliability of the proposed method.

Traffic Flow Prediction Methods Based on Deep Learning

In recent years, due to the rapid development of deep learning, it has been found that deep learning has shown good experimental results in traffic flow prediction, and is significantly better than traditional traffic flow prediction methods. This paper mainly classifies, sorts out, and summarizes the classic techniques and research status of traffic flow prediction based on deep learning. According to the length of traffic flow data, the traffic flow prediction method based on deep learning is further subdivided into long traffic flow prediction method and short traffic flow prediction method. The representative models of the two traffic flow prediction methods are analyzed and introduced in detail. Each representative model is expanded and extended, and the performance evaluation indicators of the long and short traffic flow prediction models are introduced. Finally, the possible future research directions and corresponding development trends in this field are summarized.

A Hybrid Approach to Mountain Torrent-Induced Debris Flow Prediction Combining Experiments and Gradient Boosting Regression

Debris flows are highly unpredictable and destructive natural hazards that present significant risks to both human life and infrastructure. Despite advances in machine learning techniques, current predictive models often fall short due to the insufficient and low-quality historical data available for training. In this study, we introduce a hybrid approach that combines physical model experiments with a gradient boosting regression model to enhance the accuracy and reliability of debris flow predictions. By integrating experimental data that closely simulate real-world flow conditions, the gradient boosting regression model is trained on a more robust foundation, enabling it to capture the complex dynamics of debris flows under various conditions. Selecting slide mass, slope length, yield stress, and slope angle as explanatory variables, we focus on quantify two critical debris flow parameters—frontal thickness and velocity—at indicated locations within the flow. The model demonstrates strong predictive performance in forecasting these key parameters, achieving coefficients of determination of 0.938 for slide thickness and 0.934 for slide velocity. This hybrid approach not only simplifies the prediction process but also significantly improves its precision, offering a valuable tool for real-time risk assessment and mitigation strategies in debris flow-prone regions.

Tucker Decomposition Enhanced Dynamic Graph Convolutional Networks for Crowd Flows Prediction

Crowd flows prediction is an important problem for traffic management and public safety. Graph Convolutional Network (GCN), known for its ability to effectively capture and utilize topological information, has demonstrated significant advancements in addressing this problem. However, GCN-based models were often based on predefined crowd-flow graphs via historical movement behaviors of human beings and traffic vehicles, which ignored the abnormal changes in crowd flows. In this study, we propose a multi-scale fusion GCN-based framework with Tucker decomposition named mTDNet to enhance dynamic GCN for Crowd flows prediction. Following the paradigm of extant methods, we also employ the predefined crowd-flow graphs as a part of mTDNet to effectively capture the historical movement behaviors of crowd flows. To capture the abnormal changes, we propose a Tucker decomposition-based network with the product of the adjacency matrix of historical movement pattern graphs and an adaptive learning tensor ( \(ALT\) ) by reconstructing the crowd flows. Particularly, we utilize the Tucker decomposition scheme to decompose \(ALT\) , which enhances the dynamic learning of graph structures, allowing for effective capturing of the dynamic changes in crowd flow, including abnormal changes. Furthermore, a multi-scale three-dimensional GCN is utilized to mine and fuse the multiscale spatio-temporal information from crowd flows, to further boost the mTDNet prediction performance. Experiments conducted on two real-world datasets showed that the proposed mTDNet surpasses other crowd flow prediction methods.

3D human model guided pose transfer via progressive flow prediction network

Human pose transfer is to transfer a conditional person image to a new target pose. The difficulty lies in modeling the large-scale spatial deformation from the conditional pose to the target one. However, the commonly used 2D data representations and one-step flow prediction scheme lead to unreliable deformation prediction because of the lack of 3D information guidance and the great changes in the pose transfer. Therefore, to bring the original 3D motion information into human pose transfer, we propose to simulate the generation process of real person image. We drive the 3D human model reconstructed from the conditional person image with the target pose and project it to the 2D plane. The 2D projection thereby inherits the 3D information of the poses which can guide the flow prediction. Furthermore, we propose a progressive flow prediction network consisting of two streams. One stream is to predict the flow by decomposing the complex pose transformation into multiple sub-transformations. The other is to generate the features of the target image according to the predicted flow. Besides, to enhance the reliability of the generated invisible regions, we use the target pose information which contains structural information from the flow prediction stream as the supplementary information to the feature generation. The synthesized images with accurate depth information and sharp details demonstrate the effectiveness of the proposed method.

Flow prediction and dimensionless feature analysis in orifice microchannels under cavitation conditions based on similarity criteria

A novel set of dimensionless numbers for predicting cavitation flow in gas-liquid flow through a throttling orifice is proposed. Multiple sets of cavitation data are obtained from the literature, and eight dimensionless groups are extracted using dimensional analysis. Subsequently, self-similarity among these groups is established, leading to the proposal of new dimensionless correlations (CfW and CfWd). Investigation results indicate that the established correlations between dimensionless groups can predict cavitation flow in orifices under a wide range of hydrodynamic and geometric conditions. The predictive model is validated using different sets of experimental data, improving accuracy (average relative error decreased by more than 57%) and covering more physical conditions compared to previous correlation predictions. Furthermore, dimensionless parameters with significant influence on the cavitation discharge coefficient are identified by the explanatory machine learning technique SHapley Additive exPlanations (SHAP) and sensitivity analysis. Based on these findings, this investigation contributes to a reduction in the number of variables involved in cavitation experiments or simulations.

Scaling the predictions of multiphase flow through porous media using operator learning

Operator learning promises generalized and accelerated predictions for fluid flow problems. Using non-body fitted Cartesian mesh based CFD simulation data, we explore if multiphase flow in large spatial extents can be predicted through Fourier Neural Operator (FNO) using pore-scale direct numerical simulations in smaller, representative domains. While operator learning is known to be resolution independent, its relevance for scaling up a domain for spatially emergent phenomena has not been sufficiently explored in an engineering context. Firstly, we verify the predictions from FNO using single-phase flow over complex shaped obstacles and two phase flows through an arbitrary arrangement of circular cylinders, mimicking pore microstructure in a packed bed. Further, we perform multiphase flow predictions in domains larger than those used in training sets to assess the potential of FNO for ‘domain-scaling.’ We also test if transfer learning improves the prediction on the larger domain.

Advancing Mix Design Prediction in 3D Printed Concrete: Predicting Anisotropic Compressive Strength and Slump Flow

Introducing 3D-concrete printing has started a revolution in the construction industry, presenting unique opportunities alongside undeniable challenges. Among these, the major challenge is the iterative process associated with mix design formulation, which results in significant material and time consumption. This research uses machine learning (ML) techniques such as Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), Decision Tree Regression (DTR), Gaussian Process Regression (GPR), and Artificial Neural Network (ANN) to overcome these challenges. A dataset containing 21 mix constituent features and 4 output properties (cast and printed compressive strength, and slump flow) was extracted from the literature to investigate the relationship between mix design and performance. The models were assessed using a range of evaluation metrics, including Mean Absolute Error (MAE), Root Mean Squared Error (RMSE), Mean Squared Error (MSE), and R-squared value. Gaussian Process Regression (GPR) yielded more favorable results. In the case of cast strength, GPR achieved an R2 value of 0.9069, along with RMSE, MSE, and MAE values of 13.04, 170.12, and 9.40, respectively. A similar trend was observed for printed strengths in directions 1, 2, and 3. GPR achieved R-squared values exceeding 0.91 for all directions, accompanied by significantly lower RMSE values (below 4.1). The machine learning models were also validated using four unique mix designs. These mixes were 3D printed and tested for compressive strength and slump flow. GPR's average error was 10.55%, while SVM achieved a slightly lower average error of 9.38%. Overall, this work presents a novel approach for optimizing 3D-printed concrete by enabling the prediction of slump flow and compressive strength directly from the mix design. This approach can facilitate the design and fabrication of 3D-printed concrete structures that fulfill the necessary strength and printability requirements.

Prediction of bubbly flow and flow regime development in a horizontal air-water pipe flow with a morphology-adaptive multifluid CFD model

Prediction of bubbly flow and flow regime development in a horizontal air-water pipe flow with a morphology-adaptive multifluid CFD model

Efficiency of a predictive surrogate model for hemodynamic predictions of blood flow in an idealized carotid artery stenosis

Efficiency of a predictive surrogate model for hemodynamic predictions of blood flow in an idealized carotid artery stenosis

Ants integrate proprioception as well as visual context and efference copies to make robust predictions

Forward models are mechanisms enabling an agent to predict the sensory outcomes of its actions. They can be implemented through efference copies: copies of motor signals inhibiting the expected sensory stimulation, literally canceling the perceptual outcome of the predicted action. In insects, efference copies are known to modulate optic flow detection for flight control in flies. Here we investigate whether forward models account for the detection of optic flow in walking ants, and how the latter is integrated for locomotion control. We mounted Cataglyphis velox ants in a virtual reality setup and manipulated the relationship between the ants’ movements and the optic flow perceived. Our results show that ants compute predictions of the optic flow expected according to their own movements. However, the prediction is not solely based on efference copies, but involves proprioceptive feedbacks and is fine-tuned by the panorama’s visual structure. Mismatches between prediction and perception are computed for each eye, and error signals are integrated to adjust locomotion through the modulation of internal oscillators. Our work reveals that insects’ forward models are non-trivial and compute predictions based on multimodal information.

Impact of collocation point sampling techniques on PINN performance in groundwater flow predictions

Impact of collocation point sampling techniques on PINN performance in groundwater flow predictions

Robust pore-resolved CFD through porous monoliths reconstructed by micro-computed tomography: From digitization to flow prediction

Robust pore-resolved CFD through porous monoliths reconstructed by micro-computed tomography: From digitization to flow prediction