Predicted Flow Patterns Research Articles

The Development of Machine Learning Models for Radial Compressor Monitoring With Instability Detection

Abstract Modern radial compressors are designed to operate with high performance and a wide range of applications while minimizing their environmental impact. It is necessary to study the machine near the stability limit and gain a comprehensive understanding of the fluid dynamic mechanisms that trigger the instability in the system to extend the operating range at high efficiency. Machine learning aids in developing pattern identification models for detecting compressor instability. In a prior study, a two-stage radial compressor for refrigerant gas underwent extensive simulation, capturing unsteady RANS conditions and generating a substantial dataset of pressure signals from multiple probes. Selected signals, coupled with detailed computational fluid dynamics (CFD) post-processing, revealed fluid dynamic structures near surge conditions. This study utilizes all pressure signals to assess the stage's operational state (stable, transient, or unstable/surge). An additional goal is to develop an algorithm predicting flow patterns in different stage sections with minimal pressure signals at the rotor inlet. This is a preliminary step toward the development of a smart monitoring and diagnostic model for the compressor installed in a plant. The developed model consists of three submodels, trained with CFD results obtained from unsteady Reynolds averaged Navier–Stokes (URANS) simulations. The three submodels are a regression submodule, a classification submodule, and a forecasting algorithm. The regression submodule predicts diffuser static pressure fields, the classification submodule categorizes whether the compressor is in a critical condition or not, and the forecasting algorithm predicts the inducer pressure signals in the future impeller rotations according to the actual operation history. These sub-modules work together to forecast whether the compressor, according to the operation strategy, is going into instability conditions.

Gas liquid flow pattern prediction in horizontal and slightly inclined pipes: From mechanistic modelling to machine learning

This paper investigates the prediction of two-phase gas-liquid flow regimes in both horizontal and slightly inclined pipes. For this purpose, the mechanistic model of Taitel et al. (1976) and the machine learning approach have been adopted. First, the mechanistic model was implemented, tested and optimised by introducing factors in the transition equations to determine the configuration that gives the highest prediction accuracy for a specific two-phase system for which experimental data points are available. Second, several machine learning models are trained, tested and additionally validated. This is done by splitting the experimental data set corresponding to the pipe inclination range (−10∘ to 10∘) into training, test and validation sets. The best classifier achieved an accuracy of 95.5% after the test step and up to 98.9% after the validation step. Finally, the Taitel et al. model with the optimal configuration and the best machine learning classifier (XGB classifier) are used to generate the two-dimensional flow regime map.

Cyclic flow cut-off characteristics of gas-liquid two-phase flow in pipeline-riser system and prediction of its occurring condition

In offshore oil and gas fields, the prediction of gas-liquid two-phase flow pattern is of great importance for the design and operation of pipeline-riser systems. However, no special attention was put on the mechanism of periodic flow cut-off at the riser outlet, which is an inherent characteristic of certain flow patterns directly related to operation safety, in the study of flow pattern transition. In this study, differential pressure signals are used to explore the internal correlation between the flow cut-off of gas and/or liquid phase at the riser bottom and the riser outlet. The flow patterns are classified based on continuous flow or flow cut-off at the riser outlet. Then, the flow pattern transition mechanisms are studied from the view of the condition under which flow cut-off will appear. At high gas and low liquid velocities, the mechanism can be explained by a conventional theory; while at high gas and high liquid velocities, dissipation of hydrodynamic slugs becomes the major reason for flow pattern transition; and a unified model is introduced to predict the dissipation. At low gas and high liquid velocities, the condition for gas-liquid eruption can still describe the flow pattern transition, but it is modified by the slug dissipation model. At higher pressure, the lasting time of gas cut-off at the riser elbow becomes shorter, and a threshold can be set to decide whether it can be ignored. The predicted results are in good agreement with the flow pattern maps under different pipeline structures and fluid properties, and the reason why flow cut-off disappears in the experimental loop at high pressure of 10 MPa is successfully explained.

Application of Seagull Optimization Algorithm-BP Neural Network in Oil-Water Two-Phase Flow Pattern Forecasting

With the ongoing increase in global energy demand, the significance of innovations in oil exploration and development technologies is rising, especially in relation to the development of unconventional reservoirs. The application of horizontal wells is becoming increasingly important in this particular situation. However, accurately monitoring and analyzing fluids in horizontal wells remains challenging due to the complex and fluctuating flow patterns of oil-water two-phase flow within the wellbore. Several elements, including well slope angle, flow rate, and water content, are involved. This study aimed to explore and develop an effective method for forecasting flow patterns, improving the precision of the dynamic monitoring of oil-water two-phase flow in horizontal wells. By analyzing the flow patterns in different experimental conditions, a predictive model using the SOA-BP neural network was developed, providing a scientific basis for dynamic monitoring in actual production scenarios. Initially, the simulated experiment for oil-water two-phase flow was carried out at room temperature and pressure utilizing a multiphase flow simulator. An optically transparent wellbore, with a diameter comparable to that of a real downhole well, was utilized, and No. 10 industrial white oil and tap water were employed as the experimental fluids. The experiment considered multiple contributing factors, including different well deviation, total flow, and water cut. The flow characteristics of oil and water were observed via visual monitoring and high-definition video, followed by detailed analysis. After collecting the experimental data, flow regimes for various scenarios were classified based on the established theory of oil-water two-phase flow in horizontal wells; then, detailed flow distribution diagrams were drawn. These data and diagrams presented offer a visual representation of the behavioral patterns exhibited by oil-water two-phase flow under varying situations and form the basis for subsequent model training and testing. Subsequently, based on the experimental data, this study combined the Seagull Optimization Algorithm (SOA) with a BP neural network to effectively learn and predict the experimental data. The SOA optimized the weights and biases of the BP neural network, improving the model’s convergence speed and prediction accuracy. Through rigorous training and testing, an oil-water two-phase flow pattern forecasting model was established, effectively predicting flow patterns under different well deviation, total flow, and water cut conditions. Finally, to validate the efficiency of the established model, a total of 15 data points were chosen from a sample well for validation. By comparing the flow patterns predicted by the model with actual logging data, the results indicate that the model’s accuracy in identifying flow pattern was 86.67%. This demonstrates that the flow pattern prediction model based on the SOA-BP neural network achieved a high level of accuracy under different complicated working conditions. This model effectively fulfills the requirements for dynamic monitoring in actual production. This indicates that the SOA-BP neural network-based flow pattern forecasting method is highly valuable due to its practical application value and provides an efficient technical approach for the development of unconventional reservoirs and the dynamic monitoring of horizontal wells in the future.

Three-dimensional wake transition of rectangular cylinders and temporal prediction of flow patterns based on a machine learning algorithm

This study investigates the three-dimensional (3D) wake transition in unconfined flows over rectangular cylinders using direct numerical simulation (DNS). Two different cross-sectional aspect ratios (AR) and Reynolds numbers (Re) are scrutinized: AR = 0.5 at Re = 200 and AR = 3 at Re = 600. The investigation focuses on characterizing the flow patterns and forecasting their temporal evolution utilizing the proper orthogonal decomposition (POD) technique coupled with a long short-term memory (LSTM) network. The DNS results reveal the emergence of an ordered mode A for AR = 3, attributed to the stabilizing effect of the elongated AR. On the other hand, the case with a smaller AR (= 0.5) exhibits a mode-swapping regime characterized by modes A and B's distinct and simultaneous manifestation. The spanwise wavelengths of mode A and mode B are approximately 4.7 and 1.2 D for AR = 0.5, while the spanwise wavelength of mode A is 3.5 D for AR = 3. The POD serves as a dimensionality reduction technique, and LSTM facilitates temporal prediction. This algorithm demonstrates satisfactory performance in predicting the flow patterns, including the instabilities of modes A and B, across both transverse and spanwise directions. The employed algorithm adeptly predicts the pressure time series surrounding the cylinders. The duration for training the algorithm is only about 0.5% of the time required for DNS computations. This research, for the first time, demonstrates the effectiveness of the POD–LSTM algorithm in predicting complex 3D instantaneous wake transition patterns for flow past rectangular cylinders.

Novel dimensionless predictive flow pattern map for HFOs inside microscale enhanced tubes

Models designated to predict flow patterns in microscale geometries with enhanced surfaces such as micro-finned tubes are scarce in the literature and new low-GWP HFOs could benefit from the utilization of such geometries. Therefore, HFO1234ze(E)’s condensation flow patterns were subjected to visualization inside micro-finned tubes ranging from 4 to 7 mm outer diameters with various geometrical figures. The saturation temperature was set to 30 °C, vapor qualities ranged in the scope of 0.01 to 0.9, and mass fluxes in the magnitude of 100 to 400 kg m-2 s-1. Four distinctive flow patterns are observed, namely intermittent, annular, wavy-stratified, and transitional. Stratification only transpired at low mass fluxes (mainly below 100 kg m-2 s-1) and intermittent flow was only present at vapor qualities close to full condensation. The range in which transitional flow is observable shrinks with a progressive trend of mass flux. The impact of diameter was observed to be negligible, however, a more meticulous assessment highlights smaller ranges of vapor qualities in which transitional flow is present for the tube of 5 mm OD whose helix angle is substantially larger. Datapoints were juxtaposed to previous models of Doretti et al., Chen et al., Mandhane et al., and Jige et al. and the results attested to an inadequacy of accurate predictions made for tube of 4 mm OD. Noting the absence of surface tension force in the aforementioned maps, a novel flow pattern map based on modified Froude versus modified Weber numbers provided an accurate prediction for the three cases under study. Ultimately the model was also deemed fairly suitable for visualization datasets collected from literature.

Numerical simulation of endovascular treatment options for cerebral aneurysms

AbstractPredicting the long‐term success of endovascular interventions in the clinical management of cerebral aneurysms requires detailed insight into the patient‐specific physiological conditions. In this work, we not only propose numerical representations of endovascular medical devices such as coils, flow diverters or Woven EndoBridge but also outline numerical models for the prediction of blood flow patterns in the aneurysm cavity right after a surgical intervention. Detailed knowledge about the postsurgical state then lays the basis to assess the chances of a stable occlusion of the aneurysm required for a long‐term treatment success. To this end, we propose mathematical and mechanical models of endovascular medical devices made out of thin metal wires. These can then be used for fully resolved flow simulations of the postsurgical blood flow, which in this work will be performed by means of a Lattice Boltzmann method applied to the incompressible Navier–Stokes equations and patient‐specific geometries. To probe the suitability of homogenized models, we also investigate poro‐elastic models to represent such medical devices. In particular, we examine the validity of this modeling approach for flow diverter placement across the opening of the aneurysm cavity. For both approaches, physiologically meaningful boundary conditions are provided from reduced‐order models of the vascular system. The present study demonstrates our capabilities to predict the postsurgical state and lays a solid foundation to tackle the prediction of thrombus formation and, thus, the aneurysm occlusion in a next step.

Study on acoustic characteristics of bubble interaction

Abstract Bubble flow exists widely in various hydraulic systems. The shape and size of the initial bubble deformed from a nozzle/orifice have a significant effect on the dynamic characteristics of the rising processes. In the monitoring of the sound signal of the bubble flow, it is found that the bubble will produce the sound signal when it detaches and rises, and the bubble flow pattern is related to the acoustic signal. In order to better understand the physical mechanism of passive acoustic emission, the acoustic characteristics of different nozzle arrays and gas flow rates were studied. The bubble acoustic statistical characteristics are obtained by filtering. The spectrogram based on time-frequency analysis is used to count the formation of individual bubbles and also to show the transient behaviour of bubble acoustics. Synchronous analysis is applied to study the image information and acoustic signal characteristics, the corresponding relationship between the flow pattern and the acoustic texture feature is established. This will provide a reference for predicting flow patterns through acoustic signals, and is significant for improving work efficiency and detecting faults.

Towards spatio-temporal prediction of cavitating fluid flow with graph neural networks

In this article, we present a deep learning-based surrogate model for spatio-temporal prediction of cavitating fluid flow. Specifically, we introduce a finite element-inspired rotation equivariant hypergraph neural network for inferring and predicting dynamical behaviors of cavitating flow. We generate ground-truth spatial–temporal data by simulating a full-order variational system based on homogeneous mixture-based cavitation theory. We consider the flow past a NACA0012 hydrofoil to examine the predictive ability of the proposed graph neural network for cavitation dynamics. Results demonstrate that the network achieves stabilized and accurate temporal predictions of the system states, successfully forecasting the evolution patterns of individual cavitation events. Additionally, comparisons of predicted fluid loading coefficients are in good agreements with the ground-truth values. We also discuss some challenges encountered in the long-term prediction of flow patterns across multiple cavitation events. The proposed framework has implications for design optimization, active control and development of a physics-based digital twin of a cavitating marine propeller.

A new ECT image reconstruction algorithm based on Vision transformer (ViT)

Aiming at the problem of low accuracy of ECT image reconstruction, this paper proposes an ECT image reconstruction algorithm based on Vision Transformer (ViT). ViT is a method that applies the Transformer to the field of image classification, which is characterized by strong long-distance dependency learning ability and strong multi-modal fusion ability compared to CNN. This paper fully utilizes the characteristics of ViT to transform the image reconstruction process of ECT into the classification process of ViT. Here, a method is proposed to use the object field distribution as the image classification label. This method converts a two-dimensional image of the object field distribution into a one-dimensional vector, which is the label vector of the image classification. These vectors and their corresponding reconstructed images obtained by the Landweber algorithm form the sample set of ViT. Extracting a large number of flow pattern samples with various shapes and distributions through COMSOL is used as a training set. After training ViT, this network model is used to infer the labels of the predicted flow patterns. After post-processing this label, the corresponding reconstructed image can be obtained. Finally, simulation experiments are conducted, and the experiments results show that the image errors and correlation coefficients of the reconstructed images obtained through the algorithm in this paper are better than those of Tikhonov algorithm, Landweber algorithm and Long Short-Term Memory Network(LSTM). And this algorithm has better resistance to noise interference than Tikhonov algorithm, Landweber algorithm, and LSTM. This also provides a new approach and means for ECT image reconstruction.

Study of the flow pattern and pressure drop law of the air-water-foam three-phase flow

The inaccurate prediction of flow patterns and pressures after adding different types and concentrations of surfactants to wellbores and drainage lines is a common problem in shale gas wells and tubing foam drainage. To clarify the change rule of the air–water-foam three-phase flow pattern and pressure drop after adding different types and concentrations of surfactants, a surface tension test was conducted in this study. In addition, visual air–water-foam three-phase flow indoor simulation experiments were performed with various surfactants such as cocamidopropyl hydroxysulfobetaine (MX-1), dodecyldimethyl betaine (TCJ-1), and sodium α-alkenyl sulfonate (XJHSM), surfactant concentrations (0.3–0.6 %), oil pipe diameters, pipe inclinations, gas–liquid ratios, and oil contents on large-scale experimental equipment. Based on the gas–liquid distribution characteristics, the air–water-foam three-phase flow patterns in the inclined tube were reclassified, and the quantitative conversion boundaries of the various flow patterns were determined. Based on the pressure drop pulse characteristics, characterization parameters such as the scaling factor, foaming capacity, foam density, and gas-holding rate were introduced after considering the effects of various factors on the pressure drop weights, enabling a new pressure drop calculation method for air–water-foam three-phase flows for application to different types and concentrations of surfactants. The errors were verified using data from previous studies and field measurements that were within 15% and 5%, respectively. The results of these studies provide a better understanding of the air–water-foam three-phase flow patterns and pressure drop variations in shale gas wells and gathering lines.

Prediction of two-phase flow patterns based on machine learning

Due to the advantages of flexibility, security and economy, small modular reactor (SMR) has become a research hotspot in the field of nuclear energy. Gas liquid two-phase flow is one of the most common phenomena for the research and development of SMR. An effective two-phase flow pattern prediction model with high accuracy is very crucial since it will impact the thermal–hydraulic and heat transfer phenomena, and consequently, the safety of SMR. In this paper, support vector machine (SVM), Random Forest (RF) and K-Nearest Neighbor (KNN) algorithm were chosen as the candidates of machine learning (ML) models. we selected 12 databases to train and test ML models. Through the identification of flow pattern feature correlation and the preprocess of dataset, we proposed an ML model for gas–liquid flow pattern prediction based on the improved K-Nearest Neighbor (KNN). its accuracy can reach more than 99%, higher than the traditional methods.

Experimental investigation and prediction of overall and local flow patterns in pipeline-riser systems

A unified model for predicting flow patterns in the pipeline-riser system is essential for the flow assurance of offshore oil and gas pipelines. This paper uses the visual images and pressure signals to obtain the internal correlation and the transition boundaries of the overall and local flow patterns in a 46 mm ID, 1722 m long pipeline S-shaped riser system. The flow pattern map with industrial parameters is plotted, and a fast identification method for unstable flow is proposed based on the differential pressure signal of the riser. The mapping relationship between the overall flow pattern divided by the pressure signals and the local flow pattern distinguished by the visual images is established. The transition mechanism dominated by the liquid phase is whether the gas in the downward-inclined pipeline accelerates after entering the riser base, and the transition mechanism dominated by the gas phase is whether the gas can penetrate the liquid in the riser so that the gas does not accumulate in the downward-inclined pipeline. A unified mechanistic model is proposed for predicting the overall flow pattern transition boundaries in pipeline-riser systems. The model results show a good agreement with 24 kinds of flow pattern maps under various pipeline structures, fluid properties, and separator pressures.

A Reliability-Based Network Equilibrium Model with Electric Vehicles and Gasoline Vehicles

With the popularity of electric vehicles, they have become an indispensable part of traffic flow on the road network. This paper presents a reliability-based network equilibrium model to realise the traffic flow pattern prediction on the road network with electric vehicles and gasoline vehicles, which incorporates travel time reliability, electric vehicles’ driving range and recharge requirement. The mathematical expression of reliable path travel time is derived, and the reliability-based network equilibrium model is formulated as a variational inequality problem. Then a multi-criterion labelling algorithm is proposed to solve the reliable shortest path problem, and a column-generation-based method of the successive average algorithm is proposed to solve the reliability-based network equilibrium model. The applicability and efficiency of the proposed model and algorithm are verified on the Nguyen-Dupuis network and the real road network of Sioux Falls City. The proposed model and algorithm can be extended to other road networks and help traffic managers analyse traffic conditions and make sustainable traffic policies.

Virtual Meter with Flow Pattern Recognition Using Deep Learning Neural Networks: Experiments and Analyses

Summary Operators often require real-time measurement of fluid flow rates in each well of their fields, which allows better control of production. However, petroleum is a complex multiphase mixture composed of water, gas, oil, and other sediments, which makes its flow challenging to measure and monitor. A critical issue is how the liquid component interacts with the gaseous phase, also known as the flow pattern. For example, sometimes liquids can accumulate in the lower part of the pipeline and block the flow completely, causing a gas pressure buildup that can lead to unstable flow regimes or even accidents (blowouts). On the other hand, some flow patterns can also facilitate sediment deposition, leading to obstructions and reduced production. Thus, this work aims to show that deep neural networks can act as a virtual flowmeter (VFM) using only a history of production, pressure, and temperature telemetry, accurately estimating the flow of all fluids in real time. In addition, these networks can also use the same input data to detect and recognize flow patterns that can harm the regular operation of the wells, allowing greater control without requiring additional costs or the installation of any new equipment. To demonstrate the feasibility of this approach and provide data to train the neural networks, a water-air loop was constructed to resemble an oil well. This setup featured inclined and vertical transparent pipes to generate and observe different flow patterns and sensors to record temperature, pressure, and volumetric flow rates. The results show that deep neural networks achieved up to 98% accuracy in flow pattern prediction and 1% mean absolute prediction error (MAPE) in flow rates, highlighting the capability of this technique to provide crucial insights into the behavior of multiphase flow in risers and pipelines.

Experimental investigation of two-phase flow in Chevron-type compact plate heat exchangers: A Study on pressure drops and flow regimes visualization

Compact plate heat exchangers are a very promising technology and lately they are being considered for potential use as steam generators in Small Modular Reactors. However, there is a lack of scientific literature on their operation with two-phase flows, especially with non-refrigerant fluids. In this study, we conducted experiments to visualize and measure the pressure drop of a two-phase flow in a Chevron-type Plate Heat Exchanger. An air–water mixture was used in adiabatic conditions as the operating fluid in upward-flow configuration. Visualization was achieved through high-framerate videos. This study covers a wide range of operating conditions, surpassing those documented in existing literature by specifically analyzing also the region of very low mass flux for both phases. The tested conditions ranged from 6 to 365 kg/m2s of water and from 0.02 to 5 kg/m2s of air. Single-phase pressure drops were measured to establish a correlation for the Darcy friction factor. Instead, measurements of the pressure drop in two-phase conditions were processed and presented using a non-dimensional form referred to as the two-phase multiplier. The adoption of a void-fraction model played a crucial role in accurately extrapolating the frictional component of the pressure drop from the measurements, resulting in less scattered data on the Lockhart–Martinelli plot. In addition, the observed flow structures were categorized into distinct regimes (fine-coarse bubbly, Taylor-like bubbly, heterogeneous, partial film, and film flow) based on visual observation and were represented on a flow map. Finally, these data were used to develop new criteria for predicting flow pattern transitions.

Two-Phase Flow Pattern Identification in Vertical Pipes Using Transformer Neural Networks

The oil and gas industry consistently embraces innovative technologies due to the significant expenses associated with hydrocarbon transportation, pipeline corrosion issues, and the necessity to gain a deeper understanding of two-phase flow characteristics. These solutions involve the implementation of predictive models utilizing neural networks. In this research paper, a comprehensive database comprising 4864 data points, encompassing information pertaining to oil–water two-phase flow properties within vertical pipelines, was meticulously curated. Subsequently, an encoder-only type transformer neural network (TNN) was employed to identify two-phase flow patterns. Various configurations for the TNN model were proposed, involving parameter adjustments such as the number of attention heads, activation function, dropout rate, and learning rate, with the aim of selecting the configuration yielding optimal outcomes. Following the training of the network, predictions were generated using a reserved dataset, thus facilitating the creation of flow maps depicting the patterns anticipated by the model. The developed TNN model successfully predicted 9 out of the 10 flow patterns present in the database, achieving a peak accuracy of 53.07%. Furthermore, the various predicted flow patterns exhibited an average precision of 63.21% and an average accuracy of 86.51%.

FLOW PATTERN PREDICTION IN HORIZONTAL AND INCLINED PIPES USING TREE-BASED AUTOMATED MACHINE LEARNING

In the oil and gas industry, understanding two-phase (gas-liquid) flow is pivotal, as it directly influences equipment design, quality control, and operational efficiency. Flow pattern determination is thus fundamental to industrial engineering and management. This study utilizes the Tree-based Pipeline Optimization Tool (TPOT), an Automated Machine Learning (AutoML) framework that employs genetic programming, in obtaining the best machine learning model for a provided dataset. This paper presents the design of flow pattern prediction models using the TPOT. The TPOT was applied to predict flow patterns in 2.5 cm and 5.1 cm diameter pipes, using datasets from existing literature. The datasets went through handling of imbalanced data, standardization, and one-hot encoding as data preparation techniques before being fed into TPOT. The models designed for the 2.5 cm and 5.1 cm datasets were named as FPTL_TPOT_2.5 and FPTL_TPOT_5.1, respectively. A comparative analysis of these models alongside other standard supervised machine learning models and similar state-of-the-art similar two-phase flow prediction models was carried out and the insights on the performance of these TPOT designed models were discussed. The results demonstrated that models designed with TPOT achieve remarkable accuracy, scoring 97.66% and 98.09%, for the 2.5 cm and 5.1 cm datasets respectively. Furthermore, the FPTL_TPOT_2.5 and FPTL_TPOT_5.1 models outperformed other counterpart machine learning models in terms of performance, underscoring TPOT’s effectiveness in designing machine learning models for flow pattern prediction. The findings of this research carry significant implications for enhancing efficiency and optimizing industrial processes in the oil and gas sector.

A Comparative Study of Oil–Water Two-Phase Flow Pattern Prediction Based on the GA-BP Neural Network and Random Forest Algorithm

As global oil demand continues to increase, in recent years, countries have continued to expand the development of oil reserves, highlighting the importance of oil. In order to adapt to different strata distribution conditions, domestic drilling technology is becoming more and more perfect, resulting in a gradual increase in horizontal and inclined wells. Because of the influence of various downhole factors, the flow pattern in the wellbore will be more complex. Accurately identifying the flow pattern of multiphase flow under different well deviation conditions is very important to interpreting the production log output profile accurately. At the same time, in order to keep up with the footsteps of artificial intelligence, big data and artificial intelligence algorithms are applied to the oil industry. This paper uses the GA-BP neural network and random forest algorithm to conduct fluid flow pattern prediction research on the logging data of different water cuts at different inclinations and flow rates. It compares the predicted results with experimental fluid flow patterns. Finally, we can determine the feasibility of these two algorithms for predicting flow patterns. We use the multiphase flow simulation experiment device in the experiment. During the process, the flow patterns are observed and recorded by visual inspection, and the flow pattern is distinguished by referring to the theoretical diagram of the oil-water two-phase flow pattern. The prediction results show that the accuracy of these two algorithms can reach 81.25% and 93.75%, respectively, which verifies the effectiveness of these two algorithms in the prediction of oil–water two-phase flow patterns and provides a new idea for the prediction of oil–water two-phase flow patterns and other phases.

A physics-based machine learning technique rapidly reconstructs the wall-shear stress and pressure fields in coronary arteries.

With the global rise of cardiovascular disease including atherosclerosis, there is a high demand for accurate diagnostic tools that can be used during a short consultation. In view of pathology, abnormal blood flow patterns have been demonstrated to be strong predictors of atherosclerotic lesion incidence, location, progression, and rupture. Prediction of patient-specific blood flow patterns can hence enable fast clinical diagnosis. However, the current state of art for the technique is by employing 3D-imaging-based Computational Fluid Dynamics (CFD). The high computational cost renders these methods impractical. In this work, we present a novel method to expedite the reconstruction of 3D pressure and shear stress fields using a combination of a reduced-order CFD modelling technique together with non-linear regression tools from the Machine Learning (ML) paradigm. Specifically, we develop a proof-of-concept automated pipeline that uses randomised perturbations of an atherosclerotic pig coronary artery to produce a large dataset of unique mesh geometries with variable blood flow. A total of 1,407 geometries were generated from seven reference arteries and were used to simulate blood flow using the CFD solver Abaqus. This CFD dataset was then post-processed using the mesh-domain common-base Proper Orthogonal Decomposition (cPOD) method to obtain Eigen functions and principal coefficients, the latter of which is a product of the individual mesh flow solutions with the POD Eigenvectors. Being a data-reduction method, the POD enables the data to be represented using only the ten most significant modes, which captures cumulatively greater than 95% of variance of flow features due to mesh variations. Next, the node coordinate data of the meshes were embedded in a two-dimensional coordinate system using the t-distributed Stochastic Neighbor Embedding (-SNE) algorithm. The reduced dataset for -SNE coordinates and corresponding vector of POD coefficients were then used to train a Random Forest Regressor (RFR) model. The same methodology was applied to both the volumetric pressure solution and the wall shear stress. The predicted pattern of blood pressure, and shear stress in unseen arterial geometries were compared with the ground truth CFD solutions on "unseen" meshes. The new method was able to reliably reproduce the 3D coronary artery haemodynamics in less than 10 s.