Predicted Fluid Flow Research Articles

Autoregressive transformers for data-driven spatiotemporal learning of turbulent flows

A convolutional encoder–decoder-based transformer model is proposed for autoregressively training on spatiotemporal data of turbulent flows. The prediction of future fluid flow fields is based on the previously predicted fluid flow field to ensure long-term predictions without diverging. A combination of convolutional neural networks and transformer architecture is utilized to handle both the spatial and temporal dimensions of the data. To assess the performance of the model, a priori assessments are conducted, and significant agreements are found with the ground truth data. The a posteriori predictions, which are generated after a considerable number of simulation steps, exhibit predicted variances. The autoregressive training and prediction of a posteriori states are deemed crucial steps toward the development of more complex data-driven turbulence models and simulations. The highly nonlinear and chaotic dynamics of turbulent flows can be handled by the proposed model, and accurate predictions over long time horizons can be generated. Overall, the potential of using deep learning techniques to improve the accuracy and efficiency of turbulence modeling and simulation is demonstrated by this approach. The proposed model can be further optimized and extended to incorporate additional physics and boundary conditions, paving the way for more realistic simulations of complex fluid dynamics.

3-D Simulations of Groundwater Utilization in an Urban Catchment of Berlin, Germany

Abstract. The objective of this study is to analyze the influence of groundwater pumping on predicted groundwater circulation below the urban center of Berlin, Germany, by 3-D numerical models. Of particular interest are hydraulic head distributions, the related shallow-deep groundwater interactions and their scale dependency within an anthropogenically overprinted environment. For this purpose, two model scenarios are investigated. In the first model realization (Model 1), the effects of groundwater pumping are implemented by imposing a fixed, though spatially variable, hydraulic head distribution over the whole model area, therefore implicitly taking into account the effects of pumping activities. In the second model realization (Model 2), these effects are considered in an explicit manner by imposing variable production rates in locations where pumping activities are ongoing. The results of this study show, that both models predict similar hydraulic head distributions on the regional scale (i.e. urban wide). Locally, differences in the extent, volume and depth of emerging depression cones can be observed. This is manifested in differences in predicted fluid flow patterns supporting or refuting the possibility of contaminant transport in an area of importance for groundwater production (Lower Havel). Herein, the second model approach outlines the necessity of implementing wells as an active parameter to reproduce observed fluid pathways.

A comparative fluid flow characterisation in a low frequency/high power sonoreactor and mechanical stirred vessel

This study aims at analysing the jet-like acoustic streaming generated under low-frequency and high-power ultrasound irradiation and comparing it with fluid streaming generated by traditional mechanical mixing. The main characteristics of fluid flow, which include radial, axial and tangential terms of velocity and their effects on fluid flow pattern, pressure distribution, axial mixing time and turbulence intensity were considered at different power inputs. Both 3D CFD simulation and Particle Image Velocimetry (PIV) were used in this study. The CFD results indicated that the jet-like acoustic streaming reached the velocity magnitude of 145cm/s at 400W, which reduced the mixing time to 1.38s. However, the minimum mixing time of 3.18s corresponding to the impeller rotational speed of 800RPM was observed for mechanical stirring. A uniform axial flow pattern was generated under ultrasound irradiation whereas the tangential flow pattern was more prominent in the stirred vessel. Besides, the highest turbulence was observed in the vicinity of the ultrasound transducer and impeller with the values of 138% and 82% for the ultrasonicator and stirred vessel, respectively. The predicted fluid flow pattern under ultrasound irradiation was in a reasonable agreement with that obtained from PIV, with a reasonable accuracy.

Geological structure of the Paradox Valley Region, Colorado, and relationship to seismicity induced by deep well injection

AbstractThe Bureau of Reclamation (Reclamation) operates a deep injection well at Paradox Valley in western Colorado. Seismicity has been recorded since 1985, with more than 5900 likely induced earthquakes detected since injection operations began in 1991, making this project a unique and valuable case study in the long‐term development of injection‐induced seismicity. Likely induced earthquakes occur in a complex pattern with clusters separated by aseismic gaps of 2 km or more, and have been observed up to 16 km from the injection well. To evaluate how the subsurface geologic structure may control the occurrence of induced seismicity, we compare the earthquake locations to a set of geologic models that Reclamation commissioned prior to drilling the injection well. Our analysis indicates that many aspects of the observed seismicity pattern can be attributed to the complex geologic structure. The earthquake depths and locations are generally consistent with the structural trends of the primary injection target formation, within the uncertainty of the models, and some of the mapped basement faults appear to serve as boundaries in the earthquake locations, suggesting that they are barriers to flow. Additionally, the majority of events follow the predicted fluid flow pattern, which trends northwest–southeast and wraps around Paradox Valley. Some features of the seismicity pattern, such as the early onset of seismicity to the northwest of the injection well and the distinct clustering, are not explained by the current geologic models and may be related to unmapped geologic features or local variations in the stress field.

Comparison between Mixing in Novel High Frequency Sonoreactor and Stirred Tank Reactor

This paper reports the results of the Computational Fluid Dynamics (CFD) modeling for a newly developed 1.7 MHz ultrasound reactor. No chemical reaction was studied and only mechanical mixing was discussed. The predicted fluid flow patterns in this sonoreactor was compared with that of a stirred tank reactor equipped with a classic Rushton turbine impeller. Results of model prediction illustrate significant differences in the predicted flow pattern inside the studied reactors. The higher performance of the sonoreactor was obtained in terms of the axial circulation times, which were 1.37 and 7.12 s for the sonoreactor and the stirred tank reactor, respectively. In order to validate the predicted results, they were compared with the published experimental data for both reactors, and reasonable agreements have been obtained.

Modeling of Blast Furnace with Layered Cohesive Zone

An ironmaking blast furnace (BF) is a moving bed reactor involving counter-, co-, and cross-current flows of gas, powder, liquids, and solids, coupled with heat exchange and chemical reactions. The behavior of multiple phases directly affects the stability and productivity of the furnace. In the present study, a mathematical model is proposed to describe the behavior of fluid flow, heat and mass transfer, as well as chemical reactions in a BF, in which gas, solid, and liquid phases affect each other through interaction forces, and their flows are competing for the space available. Process variables that characterize the internal furnace state, such as reduction degree, reducing gas and burden concentrations, as well as gas and condensed phase temperatures, have been described quantitatively. In particular, different treatments of the cohesive zone (CZ), i.e., layered, isotropic, and anisotropic nonlayered, are discussed, and their influence on simulation results is compared. The results show that predicted fluid flow and thermochemical phenomena within and around the CZ and in the lower part of the BF are different for different treatments. The layered CZ treatment corresponds to the layered charging of burden and naturally can predict the CZ as a gas distributor and liquid generator.

Particle fracture, retention, and fluid flow in metal matrix composite friction joints

Friction joining of metal matrix composite (MMC)/MMC and MMC/AISI 304 stainless steel base materials is examined using a combination of experimental testing and numerical modeling. In particular, the fracture of reinforcing particles during the friction-joining operation is investigated. The particle diameter and interparticle spacing decrease and the area fraction of particles markedly increases in material immediately adjacent to the bondline. Smaller particles are observed in frictionwelded joints produced using high friction pressures. The principal effect of the forging operation is in decreasing the interparticle spacing. There was excellent correspondence between predicted fluid flow in A1/A1 joints and experimental test results examining the transfer of Al2O3 particles during the alloy 6061/alloy 6061 friction-joining operation. It is suggested that small-diameter particles formed due to fracture early in the friction-joining operation are retained at the bondline of MMC/MMC joints as a direct consequence of the flow of plasticized material and reinforcing particles in the contact zone. A combination of numerical modeling of fluid flow and direct experimental testing have confirmed that Al2O3 particles transfer from the stationary to the rotating boundary in MMC/MMC friction joints. Also, limit cycles embedded within the flow favor the retention of smalldiameter fractured particles at the bondline of MMC/MMC joints. A quite different situation exists in dissimilar MMC/AISI 304 stainless steel joining. In dissimilar joints, a dynamically quiescent region is formed immediately adjacent to the stainless steel boundary. It is suggested that the absence of flow of plasticized material promotes retention of fractured alumina particles in dissimilar joints.

Simulation and measurement of enhanced turbulent heat transfer in a channel with periodic ribs on one principal wall

This paper performs a numerical and experimental analysis to investigate the heat transfer and fluid flow behavior in a rectangular channel flow with streamwise-periodic ribs mounted on one of the principal walls. The k-ε- A PDM turbulence model together with a smoothed hybrid central/skew upstream difference scheme (SCSUDS) and the PISO pressure-velocity coupling algorithm was applied to solving the accelerated, separated and recirculating flows. The real-time holographic interferometry technique was adopted to measure the time-dependent temperature field in the ribbed duct. The predicted fluid flow and temperature field were tested by previous LDV measurements and present holographic interferometry data, and reasonable agreement was achieved. By the examination of the local wall temperature distribution for the uniform wall heat flux (UHF) boundary condition the regions susceptible to the hot spots are identified. Moreover, the study provided the numerical solution to investigate the effect of geometry and flow parameters on the local as well as average heat transfer coefficients. The compact correlation of the average heat transfer coefficient was further developed and accounted for the rib height, rib spacing, and Reynolds number.

Fundamental viscous solutions or 'transient oseenlets’ associated with a body manoeuvring in a viscous fluid

A viscous boundary element method involving a convolution-integral formulation is developed to determine directly the fluid actions and velocity flow field associated with a body manoeuvring in a viscous fluid. The proposed general approach is analogous to a potential flow singularity distribution panel method with the singularity replaced by a fundamental viscous solution. From the developed nonlinear mathematical model based on an integral identity relation, it is seen that the Oseen equation, its variant or modified form is central to the theoretical development of the fundamental viscous solution. Analytical expressions are presented for the fundamental solutions (or transient oseenlets) appropriate to two-dimensional and three-dimensional steady, and unsteady (transient) manoeuvring problems involving prescribed translational, rotational and their combined motions. The examples considered, and hence the fundamental viscous solutions, relate to model towing tank experiments (i.e. steady state, rotating arm, planar motion mechanism oscillatory tests, etc.) used extensively to derive fluid action data on manoeuvring ships and submersibles. Examples of combined translational and rotational motions are investigated and these relate to an elementary idealization of a propeller rotating in an axial flow and an idealization of a body moving ahead with parasitic oscillatory roll motion. The predicted fluid flow patterns associated with the various fundamental solutions are presented and these clearly illustrate the generation, shedding and decay of vortices in the wakes.

Role of near-wall node location on the prediction of melt flow and residence time distribution in tundishes by mathematical modeling

Several researchers have been engaged in three-dimensional (3-D), mathematical modeling of melt flow in tundishes using finite difference techniques over the past 10 to 15 years. Most of these computational schemes employ thek-e model of turbulence and are capable of predicting the general flow pattern in different flow systems. However, detailed comparison of computed results with water-model experiments in the tundish system has shown that the finite difference grid spacing has an influence on the predicted fluid flow and the residence time distribution of the melt. In the present work, the effect of variation in the grid spacing on the flow predictions is presented. Further, guidelines are presented for proper spacing of grids to enhance the predictive capabilities of such models.