Classification Of Flow Patterns Research Articles

An entropy measure-based study on flow pattern of gas-liquid two-phase flow in a U-Tube

U-tubes are widely applied in gas-liquid two-phase transportation in chemical engineering. The diverse flow patterns within these tubes significantly affect the pressure loss, heat transfer efficiency, and even the fluid-induced vibration amplitude of the tubes. This study explores the complex flow pattern features in a U-tube in a vertical plane and focuses on recognizing them. For the acquisition and classification of flow patterns, a Computational Fluid Dynamics (CFD) model for gas-liquid two-phase flow is first established, and its quantitative calculation error is ensured to be less than 5%. Then, the spatiotemporal evolution characteristics of flow patterns is analyzed. The real-time pressure drop response is chosen as the representation signal, and its nonlinear features in the time and frequency domain under different flow patterns are explored. A nonlinear time series is constructed by extracting a segment from the real-time pressure drop data, and six entropy measures are applied to analyze and identify them. Finally, the sensitivity of entropy measures to both the time series lengths and the tested sections are evaluated. Results show that there are six typical flow patterns in a U-tube. According to most entropy measures, the bubble flow has the highest complexity; however, the plug flow presents the lowest complexity. In the U-bend, pressure drop signals for the bubble and annular flows show random fluctuations within a specific range, in contrast to the marked periodicity in plug flow signals, while wavy and slug flows exhibit intermittent peak values. Including the upstream and downstream straight pipes in the analysis, rather than focusing solely on the U-bend, significantly increases the complexity of the stratified, plug, and slug flows. Fuzzy entropy is an effective tool for identifying the six flow patterns, demonstrating good resilience to variations in the length of the data series. This characteristic makes it highly useful for real-time identification of flow patterns in the U-bend sections of non-transparent U-tubes, offering considerable potential in chemical equipment.

RGB Mapping: A Dynamic Approach for Flow Pattern Identification and Classification

Thermal-hydraulic modeling and simulation are heavily dependent on the knowledge of flow regimes, which sort multiphase flow patterns into classifications with static boundaries that display characteristic features of flow patterns and phase distributions. Flow regime maps heretofore have been primarily developed mainly for vertical or horizontal pipe orientations by capturing images of flows at varying conditions and assigning them to different flow regimes based on visualization studies. As such, this method is inherently subjective. Furthermore, classifying two-phase flows into regimes with static boundaries is not physical, as two-phase flows undergo gradual changes as they transition from one regime to another. Therefore, such a static approach is not only unphysical but also inaccurate, resulting in the erroneous analysis of two-phase flows. Additionally, subjective observations at such flow conditions are prone to disagreement and inconsistencies in labeling. These difficulties are exacerbated in inclined two-phase flows. In this study, a novel two-phase flow pattern map employing the dynamic red, green, and blue (RGB) color model is presented. The model contains the embedded information of three normalized time-averaged statistics from signals obtained by a dual-ring impedance meter. A database of impedance meter signals from a 25.4-mm-inner-diameter adiabatic air-water inclinable test facility is used to compute statistics for each experimental flow condition. Multiple statistical measures are evaluated, and a principal component analysis is performed for feature selection. By mapping each parameter to intensities of red, green, and blue, a dynamic RGB flow pattern map can be produced in which groups of similar colors are representative of various flow regimes and the dynamic transition characteristics between the flow regimes are characterized physically. Therefore, this new method makes viable the observation of how flow patterns gradually change with respect to angle, as well as the fuzzification or quantification of the uncertainty surrounding experimental transition boundaries.

Patterns of venous collateral development after splenic vein occlusion associated with surgical and oncological outcomes after distal pancreatectomy.

Splenic vein occlusion (SpVO) due to a pancreatic tumor may result in the development of collateral circulation and left-sided portal hypertension. This study aimed to investigate the impact of SpVO on distal pancreatectomy (DP) and provide insights about the management of such cases. This retrospective analysis included 124 patients who underwent DP from 2014 to 2022. A subgroup analysis was performed on 88 patients who underwent DP for pancreatic ductal adenocarcinoma (PDAC). SpVO was found in 26 (20.8%) patients. The patients with SpVO had significantly larger splenic volumes and lower platelet counts. Compared to the patients with patent splenic veins (SpVs), the patients with SpVO underwent significantly longer operations (p = 0.006), with a higher incidence of postoperative complications (p = 0.002). We classified the collateral routes associated with SpVO into five patterns. The most common pattern was the left gastroepiploic vein type, which was associated with a tumor of the pancreatic body. In patients with PDAC, SpVO was associated with larger tumors, microscopic vascular permeation, and peritoneal recurrence. However, the differences between overall and recurrence-free survival rates in the patients with SpVO vs those with patent SpVs were not significant. SpVO causes left-sided portal hypertension, which can be a risk for perioperative complications in DP. Operative planning based on the classification of collateral flow patterns may help prevent intraoperative congestion and perioperative complications.

Identifying steady and pulsatile two-phase flow regimes with pressure drop signals and acoustic emissions

Identifying two-phase flow regimes is vital for understanding phase-change heat and mass transfer processes. Traditionally, flow regime identification has relied on subjective flow visualization studies, which are then converted to flow regime maps applicable to specific operating conditions. However, these conventional, largely subjective, maps fall short in predicting abrupt changes in flow patterns that often occur during regular operation of vapor compression and absorption heat pumps and other thermal systems. Conventional flow regime maps are also inadequate for addressing the transient, intermittent, or periodic flow regimes that occur in boiling and condensation enhanced using acoustic and other emerging techniques. Therefore, there is a pressing need for alternative flow regime identification techniques that can adapt and reliably track rapid and local changes in two-phase hydrodynamics. Promising candidates for dynamic flow pattern classification include high-resolution pressure drop signals and acoustic emission spectra, which can provide insights into the local hydrodynamics within a flow channel. To assess the feasibility of these measurement techniques, differential pressure and condenser microphone measurements are recorded alongside high-speed videos of a two-phase flow of saturated R134a. The total mass flux and vapor quality are varied to understand the effects of phase velocity and liquid inventory on wave propagation in the channel. Additionally, forced oscillations are introduced to a steady two-phase flow to analyze their impact on flow patterns and their corresponding pressure drop and acoustic signals. Statistical analyses, including Gaussian mixture modeling, are employed to reveal characteristic pressure drop probability associated with each flow pattern, which form the basis for developing a model capable of predicting flow regimes in both steady and oscillating flows. The resulting framework introduces a new flow regime identification technique that can adapt to dynamic operating conditions, benefiting a wide range of thermal systems and phase-change processes.

Research on the Classification of Rail Transit Stations and Passenger Flow Patterns—A Case from Xi’an, China

Transit-oriented development (TOD) has been promoted and implemented worldwide through the efficient integration of rail transit stations and land use. However, the interactions between stations and the surrounding catchment areas (CAs) are characterized by different features of the built environment and regional development. Therefore, it is necessary to develop a quantitative classification method for rail transit stations based on the built environment within a CA and to identify the passenger flow characteristics of different types of stations to develop targeted planning and design policies. In this study, Line 1 and Line 2 within the third ring road of Xi’an City were taken as the objects, and a station classification system was constructed by taking station traffic levels and different building functions within the CA as the classification factors. Secondly, indicators of the built environment, such as six different types of functions, were calculated through refined vector modeling, and 30 typical stations were typologically analyzed. Furthermore, 10 typical types—totaling 11 stations—were selected for passenger flow monitoring, and the passenger flow characteristics of the different types of stations were summarized in terms of the dimensions of stations and exits. Finally, the correlations between the indicators of the built environment and passenger traffic for different functions were quantified. This study provides a basis for the future optimization of stations and the built environment, as well as station design and management.

Twin-array capacitance sensor for multi-parameter measurement of gas-solid particle flow based on BP-Adaboost

The measurement of gas-solid particle flow is a complex and error-prone task due to the intricacies of the flow process. To address the limitations associated with Electrical Capacitance Tomography (ECT) such as the complexity of hardware structure and the computational demands of imaging algorithms, this study proposes a novel TAC measurement method that utilizes the BP-Adaboost algorithm. The proposed method utilizes TAC to measure multiple parameters of the gas-solid particle flow under several working conditions and calibrate the measured flow rate with gravity sensors. At first, we conduct experiments to identify the working conditions that have the least impact on the measurements of concentration and velocity, offering valuable insights for subsequent two-phase flow measurement. The BP-Adaboost algorithm is then utilized to classify the flow pattern based on capacitance data obtained from the TAC sensor. Finally, the pattern, concentration, and velocity are used as inputs for BP-Adaboost neural network to optimize the mass flow rate and improve the measuring accuracy. Experimental results demonstrate the effectiveness of the method, with an accuracy of 83.3% for two-phase flow pattern classification and a mass flow rate error of 6.71%. Additionally, BP-Adaboost algorithm, an ensemble learning technique, effectively avoids overfitting, leading to good generalizability.

Automated fluvial hydromorphology mapping from airborne remote sensing

AbstractMapping fluvial hydromorphology is an important part of defining river habitat. Mapping via field sampling or hydraulic modeling is however time consuming, and mapping hydromorphology directly from remote sensing data may offer an efficient solution. Here, we present a system for automated classification of fluvial hydromorphology based on a deep learning classification scheme applied to aerial orthophotos. Using selected rivers in Norway, we show how surface flow patterns (smooth or rippled surfaces vs. standing waves) can be classified in imagery using a trained convolutional neural network (achieving a training and validation accuracy of >95%). We show how integration of these classified surface flow patterns with information on channel gradient, obtained from airborne topographic LiDAR data, can be used to compartmentalize the rivers into hydromorphological units (HMUs) that represent the dominant flow features. Automated classifications were broadly consistent with manual classifications that had been made in previous ground surveys, with equivalency in automated and manually derived HMU classes ranging from 61.5% to 87.7%, depending on the river stretch considered. They were found to be discharge‐dependent, showing the temporally dynamic aspect of hydromorphology. The proposed system is quick, flexible, generalizable, and provides consistent classifications free from interpretation bias. The deep learning approach used here can be customized to provide more detailed information on flow features, such as delineating between standing waves and advective diffusion of air bubbles/foam, to provide a more refined classification of surface flow patterns, and the classification approach can be further advanced by inclusion of additional remote sensing methods that provide further information on hydromorphological features.

Computational Fluid Dynamics Simulation of Oil-Water Batch Transport in a Pumpless Virtual Flow Loop

Summary Batch transportation of oil and water is a new transportation method in oil and gas gathering and transportation pipelines. Its corrosion inhibition effect has been preliminarily verified in a horizontal pipe experiment. However, achieving overall visualization in traditional loops is difficult, resulting in limited flow pattern classification and analysis of influencing factors. Combining the advantages of the traditional flow loop and the wheel flow loop, we introduce in this paper a round-head straight pipe loop and analyze the influence of key factors on the evolution of the flow pattern of the oil-water interface and the dimensionless length of the oil-water film (L~o, L~w) on the pipe wall through computational fluid dynamics (CFD) numerical simulation. The results show that the batch transportation of oil and water using the round-head straight pipe loop is more in line with the flow characteristics of oil and water two-phase flow in gathering pipelines. Three distinct three-layered flow patterns were identified, which are Flow Pattern I (oil-in-water in the upper layer, annular flow in the middle layer, and oil as the annular phase, water as the core phase, and oil-in-water in the lower layer, abbreviated as DW/O-AN-DW/O), Flow Pattern II (oil phase in the upper layer, annular flow in the middle layer, water as the annular phase, oil as the core phase, and oil in the lower layer, abbreviated as O-AN-O), and Flow Pattern III (oil phase in the upper layer, water-in-oil dispersion flow in the middle layer, and oil in the lower layer, abbreviated as O-DO/W-O). Additionally, parametric analysis reveals that the velocity of the rigid body (ν) has the greatest influence on the coverage rate of the oil film on the pipe wall, followed by the viscosity of crude oil. The density of crude oil has the least influence. The round-head straight pipe loop model offers an accurate simulation of the process of oil and water batch transportation in actual production pipelines. Therefore, the corrosion mitigation efficiency increases with the increase in oil viscosity when the viscosity of the oil lies within the range of 0.01–1 Pa·s. This increase is due to the formation of a more stable oil film on the pipe wall at higher viscosities. When the speed of the rigid body ranges from 0.5 to 1 m/s, due to the small fluid velocity, the erosion effect on the oil film on the pipe wall is relatively small, and the corrosion mitigation efficiency remains stable within a wide range.

Experimental investigation and analysis of two-phase flow instability of flow boiling in a mini-channel heat sink

High heat flux and good axial temperature uniformity of channel flow boiling demonstrate that it is ideally suited for next-generation cooling systems. Such systems, however, are hampered by undesirable flow fluctuations, making channel flow boiling ineffective due to the rapid bubble growth within the channels. Furthermore, especially in the mini-channel heat sink, the interactions between multiple parallel channels and inlet/outlet plenum induce complex hydraulic and thermal oscillations. Therefore, lots of consistent and detailed experimental data are still needed to design a reliable flow boiling system. However, only limited studies have examined the effects of channel interaction on flow boiling instability. Thus, to provide a better understanding of the flow boiling instabilities, a mini-channel heat sink was tested with near-saturated inlet conditions using FC-72 as the working fluid. The mini-channel heat sink consisted of a 150 mm long and 70.4 mm wide base area, having 22 identical rectangular channels measuring 1.6 × 4.8 mm2. The flow instabilities were analyzed by observing the vapor backflow and flow fluctuation in mini-channels with variations in the mass velocity, pressure, and heat fluxes. The analysis is based on the detailed flow visualization in the parallel channels and inlet plenum. It was found that the rapid bubble generation in individual channels promotes vapor backflow and hydraulic oscillations. Further observation of the high-speed camera images shows that the vapor crossing over between parallel channels through the inlet plenum led to the synchronization of vapor oscillations, which is responsible for the parallel channel flow instability. This visualization study revealed six dominant flow patterns showing multi-channel flow instability. Based on the classification of dominant flow patterns, a better understanding of the operating conditions for stable flow boiling was obtained.

Analysis of Flow Characteristics around a Square Cylinder with Boundary Constraint

Based on the two-dimensional hydrodynamic model of the finite volume method and structured multigrid, the flow characteristics around a square cylinder with boundary constraint are analysed. The gap ratio G/D (G is the distance from the cylinder to the channel boundary, and D is the side length of the square cylinder) does not change the four flow patterns. Under the laminar vortex street phase, the boundary constraint only reduces the scale of the vortices. The vortex centres are pressed toward the boundaries of the channel, and a low velocity zone is formed near the boundary, but the law of vorticity attenuation along the flow direction is not changed. The flow pattern classification map shows that the boundary constraint increases the Reynolds number required to generate the turbulence flow pattern, and the range of the Reynolds number in the flow pattern of the laminar vortex street has a maximum increase range. The correlation between the time-averaged drag coefficient or the vortex shedding frequency and Reynolds number under different gap ratios indicates that the resistance of the square cylinder and the vortex shedding frequency increase accordingly with the strengthening of the boundary constraint. When G/D < 3.5, the increase is particularly obvious. Meanwhile, the correlation characteristics between the resistance or the vortex shedding frequency of the square cylinder and the Reynolds number are unrelated to the boundary constraint strength.

A hybrid model to predict the pressure gradient for the liquid-liquid flow in both horizontal and inclined pipes for unknown flow patterns

Accurate prediction of the pressure gradient (PG) for the oil-water flow requires identification of the flow pattern (FP), which is usually achieved by using either an expensive measurement system or time-consuming manual observations. This study proposes a hybrid scheme where two machine-learning (ML) models are coupled in a series to predict the PG value without any conclusive FP information. The first model (M1) determines the oil-water FP, whereas the second model (M2) predicts the oil-water PG. 1637 experimental data points for the oil-water flow in both horizontal and inclined pipes are used to develop the models. The important feature subset is identified using the modified Binary Grey Wolf Optimization Particle Swarm Optimization (BGWOPSO) algorithm. The MLs' performance is evaluated using metrics including accuracy, sensitivity, specificity, and F1-score for the M1, and coefficient of variation of root mean squared error, mean absolute percentage error (MAPE), and median absolute percentage error for the M2. The evaluation metrics are cross-validated using a repeated train-test split strategy. The results showed that the overall FP classification accuracy is greater than 91%, with 90.61% sensitivity and 98.53% specificity using the weighted majority voting for M1. With the Gaussian Process regression for M2, the evaluation metrics for the PG prediction were found to be 10.65%, 86.26 Pa/m, and 0.96 for MAPE, root mean square error, and adjusted coefficient of determination, respectively. Statistical analysis showed that the selected features for liquids' and pipe's properties using the BGWOPSO algorithm were adequate to attain superior performance for both models. The achieved MAPE using the proposed hybrid model is superior to existing mechanistic or correlation-based models reported in the literature (between 26 and 69%). The proposed hybrid scheme can significantly reduce the costs associated with identifying the oil-water flow profile and be critical in designing energy-efficient transportation of liquid-liquid flow.

Cavitation inception and evolution in cavitation on a chip devices at low upstream pressures

The concept of “hydrodynamic cavitation on a chip” offers facile generation of cavitating flows in microdomains, which can be easily scaled up by arranging short microchannels (micro-orifices) in cascade formations. In this regard, microscale cavitation in an energy-efficient test rig has the potential of increasing utilization possibilities of cavitation in a wide range of applications such as liquid-phase exfoliation. In this study, a new experimental test rig was constructed to generate microscale hydrodynamic cavitation. This setup enables cavitation bubble generation at low upstream pressures through the control of the downstream pressure of the device. Particular attention was directed to the classification of flow patterns, scale effects, and cavitating flow evolutions with an in-depth categorization of underlying mechanisms such as Kelvin–Helmholtz instability. Cavitation inception appeared in the form of a single bubble. The appearance of different attached cavitating flow patterns within the microfluidic device was accompanied by new physics, which revealed that cavitation generation and development are affected by the existence of various fluid flow phenomena, particularly the jet flow. The outcome of this study makes hydrodynamic cavitation on a chip attractive for applications, where the cavitation effects are sought in the presence of multiphase fluid flows.

Soft measurement of oil–water two-phase flow using a multi-task sequence-based CapsNet

Flow parameters measurement facilitates the understanding of two-phase flow. Due to the changeable structures of the flow, the prediction of superficial velocity of oil–water two-phase flow in large diameter pipes is still a challenging problem. Therefore, we first conducted a vertical upward oil–water two-phase flow experiment in a 125 mm ID pipe, and obtained the response signals under different flow conditions by a vertical multi-electrode array (VMEA) conductance sensor. Then, new data pre-processing (1D to 2D) techniques and information fusion techniques (network channels) are employed. Moreover, the front-end structure of the network is optimized using a combination of attention block and residual structure, and the middle structure is optimized using inception block; on the other hand, the back-end structure of the original capsule network is innovatively changed so that it can handle both the flow pattern classification and superficial velocity prediction tasks. The dynamic routing algorithm has also been improved to speed up model training. Extensive experiments validate the effectiveness of the improved modules. Finally, we compare the proposed network with its variants and other competing networks. The better performance results show that our multi-task sequence-based CapsNet has great potential for dealing with high-dimensional, time-varying and nonlinear problems in multiphase flow.

Classification of spatial-temporal flow patterns in a low Re wake based on the recurrent trajectory clustering

Coherent structures are ubiquitous in unsteady flows. They can be regarded as certain kinds of spatial-temporal patterns that interact with the neighboring field. Although they play a key role in convection and mixing, there is no consensus on how to define them, and their dynamics are complicated. In the past decades, many methods are developed to identify coherent structures based on instantaneous velocity fields (e.g., vortex identification) or long-time statistics (e.g., proper orthogonal decomposition), but the evolution process of individual structures is not well considered in the identification. In this paper, we propose a new method to classify coherent motions according to their evolution dynamics. Specifically, the evolutions are represented by trajectories in the phase space. We define a distance between two trajectories and use it to construct a network that characterizes all evolution patterns. Using spectrum clustering, we categorize these patterns into various groups. This method is applied to a low Reynolds number wake flow downstream of two cylinders-in-tandem, where one of the cylinders oscillates in the transverse direction. The flow is quasi-periodic, and four types of recurrent spatial-temporal patterns can be identified. It is a useful tool to investigate low Reynolds number unsteady flows.

Improved Calculation Method for Siphon Drainage with Extended Horizontal Sections

Slope siphon drainage is a convenient and efficient above-ground drainage method that is free of manual power and can effectively maintain the stability of potential landslides and prevent the loss of life and property. The complex engineering topography inevitably requires the use of siphon drains with a total length of more than 150 m and a horizontal section length of more than 80 m, which significantly increases the difficulty of calculating the drainage capacity and thus affects the actual utilization of the project. The traditional siphon flow rate equation does not apply to long-pipe siphon conditions, especially when the lift is close to the limit, and there are significant errors in the calculation results, for which we propose a new calculation method. The proposed method considers both air release and flow-pattern classification. Thirty-six sets of experiments were conducted to validate our proposed calculation method. The results showed that our method not only calculated the siphon flow velocity well but also predicted the main flow pattern in the siphon in the experiment well. Furthermore, the equation for calculating the siphon flow velocity was extended to the siphon operation mode with long horizontal sections.

Two-phase flow characterization through recurrence quantification analysis of the dominant features of experimental dynamics

This paper describes a novel approach to the analysis of the experimental void fraction time series detected in air–water upward two-phase flows within a vertical channel. Singular Value Decomposition (SVD) is applied to the experimental void fraction time series in order to assess a novel state space spanned by the principal components of the flow dynamics; in fact, this was demonstrated in a previous study to be effective in separating the dominant features from noise-like dynamics and, above all, in evidencing the existence of recurrent behaviours.As recurrence is a typical and fundamental signature of low-order deterministic nonlinear dynamics, the present study aims at reaching a detailed insight on the recurrent structures that characterize the experimental void fraction time series, which were measured during an experimental campaign encompassing the entire spectrum of patterns in vertical air–water upward flows.The main novelty of the present study is the adoption of the tools collectively known as Recurrence Quantification Analysis not directly to the experimental time series, as it is increasingly proposed in the analysis of two-phase flows, but to their most important principal components. The goal is to obtain a more reliable characterization of two-phase flow patterns, which is based only on the relevant features and, hence, is unbiased by noisy high-order dynamics.Reported results show that the present approach indeed provides a powerful tool for the characterisation of the variety of complex dynamics exhibited by two-phase flows, as well as for flow pattern classification.

Image identification for two-phase flow patterns based on CNN algorithms

Flow patterns are essential and useful to model the interfacial structures and heat transfer in gas-liquid two-phase flow. However, the current two-phase flow patterns classification methods mostly depend on direct visual observation. This study adopted a new flow pattern classification method based on convolutional neural network (CNN) algorithms to achieve an automatic and objective identification of two-phase flow patterns. A database of 696 test conditions, including 105642 condensing flow pattern images of methane and tetrafluoromethane in a horizontal circular tube, is collected as the input of the data-driven algorithms. After 80% of image data is fed to train and fit the parameters in the algorithms, the trained models with acceptable universality are obtained to identify five flow patterns: annular flow, bubbly flow, churn flow, slug flow and stratified flow. Compared with the manual classification, the proposed method can accurately predict two-phase flow patterns with a prediction accuracy of more than 90.63% and 91.45% for the test dataset and the entire database, respectively. The average accuracy for predicting all data points in the database is more than 97.56%. The results showed that using images as input, CNN algorithms can provide objective prediction with satisfactory accuracy and universality for two-phase flow pattern identification.

Machine Learning Algorithms for Flow Pattern Classification in Pulsating Heat Pipes

Owing to their simple construction, cost effectiveness, and high thermal efficiency, pulsating heat pipes (PHPs) are growing in popularity as cooling devices for electronic equipment. While PHPs can be very resilient as passive cooling systems, their operation relies on the establishment and persistence of slug/plug flow as the dominant flow regime. It is, therefore, paramount to predict the flow regime accurately as a function of various operating parameters and design geometry. Flow pattern maps that capture flow regimes as a function of nondimensional numbers (e.g., Froude, Weber, and Bond numbers) have been proposed in the literature. However, the prediction of flow patterns based on deterministic models is a challenging task that relies on the ability of explaining the very complex underlying phenomena or the ability to measure parameters, such as the bubble acceleration, which are very difficult to know beforehand. In contrast, machine learning algorithms require limited a priori knowledge of the system and offer an alternative approach for classifying flow regimes. In this work, experimental data collected for two working fluids (ethanol and FC-72) in a PHP at different gravity and power input levels, were used to train three different classification algorithms (namely K-nearest neighbors, random forest, and multilayer perceptron). The data were previously labeled via visual classification using the experimental results. A comparison of the resulting classification accuracy was carried out via confusion matrices and calculation of accuracy scores. The algorithm presenting the highest classification performance was selected for the development of a flow pattern map, which accurately indicated the flow pattern transition boundaries between slug/plug and annular flows. Results indicate that, once experimental data are available, the proposed machine learning approach could help in reducing the uncertainty in the classification of flow patterns and improve the predictions of the flow regimes.

Dimensionless Artificial Intelligence-Based Model for Multiphase Flow Pattern Recognition in Horizontal Pipe

Summary Multiphase flow analysis attracts a lot of attention from researchers from diverse disciplines. There are several studies including experimental, theoretical modeling, and numerical analysis that were carried out to investigate the multiphase flow. However, many facets of multiphase flow are still unresolved owing to the extremely complex nature of the multiphase flow. The complex interactions of the different phases are leading to different flow regimes that are difficult to predict but essential for developing the computational model. The identification of the flow pattern is still a challenging task. One of the growing fields is the machine learning approach, which can address such complex problems. This study aims to use machine learning to develop models that can identify the flow patterns in multiphase flow. To achieve the objective, a large set of experimental data was collected. The effect of fluid properties, such as density, viscosity, and surface tension, on the flow pattern was introduced by changing the fluid properties. The wide range of data was processed by applying a machine learning technique for predicting the flow regimes. The models were built using dimensionless parameters to extend their validity for various design and operational conditions. This approach enables to capture the main flow pattern as well as subcategories of flow patterns in the horizontal pipe. Comparison and analyses of the different machine learning tools were carried out to investigate classification of multiphase flow patterns. The results showed that different artificial intelligence (AI) methods can predict the flow pattern in horizontal pipes with high accuracy. The results of using Reynold’s number for liquid (ReL) and gas (ReG) as an input to predict the flow patterns are deficient in accuracy for the support vector machine (SVM) and discriminant analysis (DA). However, the prediction capability of the model was improved by introducing Weber’s number for liquid (WeL) and gas (WeG) along with the Reynolds numbers (ReL and ReG). The improvement in the flow pattern prediction owing to the introduction of Weber’s number is speculated because of the capturing hydrodynamic phenomenon (inertia and surface tension) owing to change in fluid properties. It infers that capturing hydrodynamic phenomena affecting the flow pattern and their transition is essential for the prediction of flow patterns in multiphase flow.

Two-phase flow pattern classification based on void fraction time series and machine learning

Two-phase flow is a complex phenomenon present in several industrial applications such as chemical reactors, power generation, and in the exploration, production, and transport of oil and natural gas. The classification of the flow pattern is a fundamental step in such applications, as it influences several derived parameters and sub-processes such as flow rate, void fraction, and pressure drop estimation. In this paper, we propose an objective approach for classifying flow patterns using time series of void fraction (from a wire-mesh sensor), signal processing and machine learning. As novel approach, the time series is modeled as a stochastic process of independent and identically distributed samples with probability density function described by a Gaussian mixture model. The estimated parameters of the mixture are then fed into a Support Vector Machine (SVM), yielding the flow pattern classification. Tests were performed with a vertical liquid–gas flow database from a 52.3-mm-diameter pipe and the results indicate a great potential for application in real systems. The average accuracy and F-score obtained was higher than 0.94 for different test sets, with standard deviation lower than 0.08 for accuracy a lower than 0.11 for F-Score, demonstrating the efficiency and generalization of the proposed method.