Thiele, Christopher (Rice University) | Araya-Polo, Mauricio (Shell International Exploration & Production, Inc.) | Alpak, Faruk Omer (Shell International Exploration & Production, Inc.) | Riviere, Beatrice (Rice University)
Direct numerical simulation of multiphase pore-scale flow is a computationally demanding task with strong requirements on time-to-solution for the prediction of relative permeabilities. In this paper, we describe the hybrid-parallel implementation of a two-phase two-component incompressible flow simulator using MPI, OpenMP, and general-purpose graphics processing units (GPUs), and we analyze its computational performance. In particular, we evaluate the parallel performance of GPU-based iterative linear solvers for this application, and we compare them to CPUbased implementations of the same solver algorithms. Simulations on real-life Berea sandstone micro-CT images are used to assess the strong scalability and computational performance of the different solver implementations and their effect on time-to-solution. Additionally, we use a Poisson problem to further characterize achievable strong and weak scalability of the GPU-based solvers in reproducible experiments. Our experiments show that GPU-based iterative solvers can greatly reduce time-to-solution in complex pore-scale simulations. On the other hand, strong scalability is currently limited by the unbalanced computing capacities of the host and the GPUs. The experiments with the Poisson problem indicate that GPU-based iterative solvers are efficient when weak scalability is desired. Our findings show that proper utilization of GPUs can help to make our two-phase pore-scale flow simulation computationally feasible in existing workflows.
In this work, the scalability of the Algebraic Multiscale Solver (AMS) (Wang et al. 2014) for the pressure equation arising from incompressible flow in heterogeneous porous media is investigated on the GPU massively parallel architecture. The solvers robustness and scalability is compared against its carefully optimized implementation on the shared-memory multi-core architecture (Manea et al. 2016), which this work is directly extending. Although several components in the AMS algorithm are directly parallelizable, its scalability on GPU's depends heavily on the underlying algorithmic details and data-structures design of each step, where one needs to ensure favorable control-and data-flow on the GPU, while extracting enough parallel work for a massively parallel environment. In addition, the type of the algorithm chosen for each step greatly influences the overall robustness of the solver. Taking all these constraints into account, we have developed a GPU-based AMS that exploits the parallelism in every module of the solver, including both the setup phase and the solution phase. The performance of AMS--with our carefully optimized algorithmic choices on the GPU for both the setup phase and the solution phase, is demonstrated using highly heterogeneous 3D problems derived from the SPE10 Benchmark (Christie et al. 2001). Those problems range in size from millions to tens of millions of cells. The GPU implementation is benchmarked on a massively parallel architecture consisting of NVIDIA Kepler K80 GPU's, where its performance is compared to the multi-core CPU architecture using an optimized multi-core AMS implementation (Manea et al. 2016) running on a shared memory multi-core architecture consisting of two packages of Intel's Haswell-EP Xeon(R) CPU E5-2667. While the GPU-based AMS parallel implementation shows good scalability for the solution stage, its setup stage is less efficient compared to the CPU, mainly due to the dependence on a QR-based basis functions solver.
Assaad, Wissam (Shell Global Solutions International B. V.) | Crescenzo, Daniele Di (Shell Exploration & Production Company) | Murphy, Darren (Shell Exploration & Production Company) | Boyd, John (Shell Exploration & Production Company)
This paper presents a method of modelling surge pressures and wave propagation that can occur during well execution. The surge pressures have an impact on formations i.e. formation fracture resulting in mud losses and non-productive time. Knowing the amplitude of pressure surges in advance can lead to operation redesign to avoid losses. Pressure waves can occur at numerous points during well execution. For example, during liner operations, pressure waves can occur dart landing or plug shearing, liner hanger setting or clearing a plugged shoetrack component. It is possible that these pressure waves can create fractures in shale and sand layers i.e. when pressure wave amplitude exceeds formation fracturing limit.
A physical model is built to compute pressure wave propagation through drill string, casing and open hole, to predict amplitude of pressure wave and to warn when a fracture may occur in formation to avoid mud losses and non-productive time.
In the model, the continuity and energy partial differential equations are built for a cylindrical fluid element contained in an elastic hollow cylinder. Method of characteristic is applied to transfer the partial differential equations into ordinary differential equations. The ordinary differential equations are solved numerically to compute pressure distribution along well depth and in time. The physical model is implemented as a Graphical User Interface (GUI) tool to be used by drilling engineers at design phase of well to avoid losses. To date it has been used for cementing and perforating operations. Pressure wave computations are performed with the model for a field in Gulf of Mexico where mud losses have occurred, and results are presented in this paper.
Fallah, AmirHossein (The University of Texas at Austin) | Gu, Qifan (The University of Texas at Austin) | Ma, Zheren (The University of Texas at Austin, now with Quantum Reservoir Impact) | Karimi Vajargah, Ali (The University of Texas at Austin, now with Quantum Reservoir Impact) | Chen, Dongmei (The University of Texas at Austin) | Ashok, Pradeepkumar (The University of Texas at Austin) | van Oort, Eric (The University of Texas at Austin) | May, Roland (Baker Hughes, a GE company)
There is a growing need for comprehensive multi-phase hydraulic models that can accurately model more complex well control situations associated with the use of Managed Pressure Drilling (MPD) techniques, complex well geometries, High-Pressure High-Temperature (HPHT) conditions, riser gas unloading, etc.
A new thermal model integrated with previously developed multi-phase hydraulics software is presented here to address this need. This de-coupled thermal model is added to a sophisticated multi-phase flow code to estimate the mud temperature in the drillstring and the annulus and in the formation adjacent to the well during complex well control situations. The model uses an explicit finite volume approach and solves the mixture energy equation for the wellbore fluids, assuming that all the phases are at thermal equilibrium. Heat transfer between the drillstring and the wellbore fluid, and between wellbore and formation is calculated using a thermal resistance network. Axial heat conduction in the mud and heat generation (e.g. at the bit) are accounted for. The steady-state results of the proposed thermal model are compared to the steady-state Hasan and Kabir model and commercial software. In addition, the transient, time-dependent temperature behavior during mud circulation is compared against the results of the commercial software. Results show a very good match for both steady-state and transient cases.
Kick scenarios are simulated to show the importance of accurate temperature estimation of the drillstring and annulus fluids in HPHT conditions. Using advanced numerical schemes, a comprehensive model for heat transfer and energy storage in combination with a user-friendly Graphical User Interface (GUI) makes this model a robust tool for estimating the transient temperature profile of the mud and the formation. The model allows for evaluation of crucial parameters during well control, such as the wellbore pressure and temperature profiles, increased outflow and pit gain during kicks, gas thermodynamic behavior including solubility and unloading at low pressure conditions, gas rising velocity, and even temperature-dependent formation strength. These added features provided by the model come without loss of previous modeling capabilities, such as accounting for area discontinuity in the well and drillstring, gas dissolution in mud, non-Newtonian fluid rheology, MPD techniques, and arbitrary 3-D well trajectories.
Details of the new model and the simulation approach are shared, and various applications of the new thermal modeling capability are illustrated.
Ice-structure interaction (ISI) is a complex process, which requires a thorough understanding of the underlying physics to ensure safe operations in the ice-covered regions. Application of discrete element method (DEM) to compute ice loads on structures is a widely accepted approach, where the equations of rigid body motions are solved for all ice pieces in the computational domain. In most ISI simulations, the ice zone is assumed to be resting on a static water foundation omitting the hydrodynamic effects (added mass, water drag, wave damping) of the interacting bodies. This assumption can introduce erroneous results to simulations of the floating ice floes behavior, which in turn will incur uncertainties in planning ice management activities.
In this paper, a smooth particle hydrodynamics (SPH) based computational fluid dynamics (CFD) code is coupled with a three-dimensional DEM model to take the hydrodynamic effects of the interacting bodies including the ice pieces into account. The ice zone is modeled as discrete elements, which allows computing interaction forces by considering contact laws. The water foundation is modeled using smooth particles, which are modelled with the Naiver-Stokes equations.
Several applications of ship and offshore structures interacting with level ice and pack ice are simulated. A scenario of an offshore supply vessel operating in the marginal ice zone (MIZ) that is subject to wave forces is also simulated to show how this approach can be used for modelling complex real-world problems. This scenario is unique in a sense that it yields a multi-physics solution, where ice-structure-wave are all included in a single CFD simulation as a fully coupled analysis. The cost of the simulation is significantly reduced by running the computations on a Graphics Processing Unit (GPU) instead of a typical CPU workstation. Some of the initial results of ice-structure interactions are presented in this paper and a reasonable agreement with reduced scale model test results are found.
Liu, H. Y. (University of Tasmania) | Fukuda, D. (University of Tasmania / Hokkaido University) | Mohammadnejad, M. (University of Tasmania) | Han, Haoyu (University of Tasmania) | Chan, Andrew H. C. (University of Tasmania)
Combined finite-discrete element method has become one of the most powerful numerical methods for modelling rock failure process in recent decades. However, most of studies focus on two-dimensional combined finite-discrete element modelling of the rock failure process. This paper further develops a hybrid finite-discrete element method proposed early by the authors for three-dimensional modelling of the rock failure processes in Brazilian tests and uniaxial compression test. The further developed three-dimensional hybrid finite-discrete element method is then parallelized using compute unified device architecture - based general purpose graphic processing unit parallel method to conduct a full-scale three-dimensional modelling of rock spalling failure process in the single Hopkinson pressure bar test. It is concluded that the three-dimensional hybrid finite-discrete element method provides a valuable numerical tools for modelling rock fracture and fragmentation and the parallelization makes it possible to be applied in the large-scale rock mass instability engineering application.
The study on rock failure process has been a challenging but hot topic since rock fracture has applications in not only breaking the rock mass for extracting valuable natural resources in mining, geothermal, and oil &; gas industries but also preventing geotechnical engineering structures such as tunnels, slopes and dams from failure and collapse. In recent decades, numerical method has been one of the most powerful tools for studying rock failure process and the combined finite-discrete element method initially proposed by Munjiza (2004) has become one of the most powerful numerical methods for modelling the rock failure process. Compared with the finite element method, the combined finite-discrete element method is more robust in modelling rock failure, especially fracture, fragmentation, and fragment movements resulting in tertiary fractures. Compared with the discrete element method, the combined finite-discrete element method is more versatile in dealing with irregular-shaped, deformable and breakable particles. However, most of studies in literatures focus on modelling the rock failure process using two-dimensional (2D) finite-discrete element methods (Mahabadi et al., 2010; Liu, 2013; Lisjak et al., 2014; Liu et al., 2015 and 2016; Mahabadi et al., 2016; An et al., 2017). Thanks to the rapid development of computing power, interactive computer graphics and topological data structure, three-dimensional (3D) finite-discrete element modelling of the rock failure process has attracted the attention of more and more researchers. Rougier et al. (2014) simulated the dynamic rock failure process in dynamic Brazilian test using a 3D combined finite-discrete element method, i.e. the so-called MUNROU (Munjiza-Rougier) code running on a supercomputer with a few hundreds of CPUs at Los Alamos National Laboratory. Mahabadi et al. (2014) implemented a 3D combined finite-discrete element method to investigate the rock failure process in Brazilian disc test and uniaxial compression test although their 3D modelling of the uniaxial compression test is far from satisfactory. Hamdi et al. (2014) simulated the complete 3D fracture process during conventional laboratory testing including Brazilian indirect tension and uniaxial and biaxial compression using a combined finite-discrete element method called ELFEN developed Rockfield Ltd. In this study, a hybrid finite-discrete element method proposed by Liu et al. (2015) on the basis of Munjiza’s (2004) open-source combined finite-discrete element libraries are further developed for three-dimensional modelling of the rock failure processes in Brazilian tests and uniaxial compression test, which extends a recent study on the 3D hybrid finite-discrete element modelling conducted by the authors (Liu et al., 2018). Moreover, the further developed 3D hybrid finite-discrete element method is parallelized using the GPGPU (general purpose graphic processing unit) parallel method initially implemented in the DFPA (dynamic failure process analysis) code (Fukuda et al., 2016) to conduct a full-scale 3D modelling of the single Hopkinson pressure bar test on the rock spalling failure process. Unlike Rougier et al.’s (2014) and probably Hamdi et al.’s (2014) (although unclear since not stated in their paper) modellings completed in the supercomputer with hundreds of CPUs, all of 3D modellings reported in this paper are completed in PC although the rock spalling test is modelled using a PC with a powerful GPU.
In the current study, a simulation of the interaction between the three-dimensional dam-break wave and the vertical square column is carried out by using the MPSGPU-SJTU solver. The simulation conditions are arranged according to the experiments performed by Yeh and Petroff (2006). The results of GPU solver are compared to other researches. The evolution procedure of three-dimensional dam-break wave, including the climb, fragmentation and rollover of free surface is presented in this paper. In the process of dam-break wave and vertical square column interaction, the net force exerted on the column is monitored and in good agreement with existing experimental data. A remarkable speedup is obtained by comparing the calculation time of the GPU solver with that of the CPU version. The effect of bottom water layer is investigated. The result shows a significant difference between flow phenomenon with and without water layer.
The impact of waves on structures is an important problem in ship and ocean engineering, including nonlinear wave surface evolution, wave climbing and slapping on structures, and severe deformation or even fragmentation of free surface under the effect of structures. In recent years, the mesh-free method MPS has gained popularity for modeling free surface flows, and it has become an alternative to traditional mesh-based methods for modeling waves. Owing to the Lagrangian nature of the mesh-free method, there is no need to deal with the free surface when it is applied to simulate nonlinear free surface flows, especially when the surface tension is not important. This property makes it particularly attractive to modeling water waves, e.g., dam-break (Zhang et al., 2011), sloshing (Yang et al., 2015), water entry (Chen et al., 2017).
The earlier MPS method was limited to the two-dimensional flow problem. This is because of the large amount of calculation of MPS method, the calculation of three-dimensional problem requires a large number of particles. In order to improve the efficiency of MPS method, researchers have two main ideas: one is the method of local encryption of particles, using fewer particles to obtain better simulation results, such as multi-resolution particle method (Tang et al., 2016), overlapping particle method (Shibata et al., 2012). Another kind of parallel algorithm is divided into two kinds from the hardware environment: one is the parallel method based on CPU environment (Ikari and Gotoh, 2008, Iribe et al., 2010), the other is the parallel method based on GPU. Zhu et al. (2011) developed different versions of MPS code based on different GPU memories. Hori et al. (2011) used CUDA (Compute Unified Device Architecture) language to develop a GPU-accelerated MPS code and only acquired about 3-7 acceleration ratio by simulating two-dimensional (2-D) dam break. Li et al. (2015) applied GPU acceleration technique to two parts of MPS, neighbor particle list and pressure Poisson equation. By simulating 3-D dam break and sloshing, the speedup of these two parts is about 1.5 and 10, respectively. Gou et al. (2016) used GPU accelerated MPS to simulate the isothermal multi-phase fuel-coolant interaction.
In this paper, we present the application of the GPU-based particle simulation to Three-dimensional (3D) complicated fluid flow problems including free surfaces with surface tension and drag force. The particle approach is based on the SPH (Smoothed Particle Hydrodynamics) method using quintic spline kernel functions. We adopt the inter-particle potential force model with a potential coefficient as a surface tension model. The GPU-implementation consists of the search for neighboring particles in the locally uniform grid cell using hash function. Numerical results demonstrate the workability and validity of the present approach through the dam-breaking flow problem, the droplet oscillation and the droplet-falling impact with surface tension and drag force.
The numerical simulations of three-dimensional (3D) viscous fluid flows including multi-scale/physics and moving boundary/obstacle are indispensable in science and engineering fields from a practical point of view. Numerical difficulties have been experienced in the solution of the Navier-Stokes equations at higher Reynolds numbers. In particular, it is well known that the centered finite difference and standard Galerkin finite element formulations lead to spurious oscillatory solutions for flow problem at high Reynolds number regimes. To overcome such spurious oscillations, various upwind/upstream-based schemes have been significantly presented by many researchers in both frameworks. On the other hand, there are various gridless/meshless-based particle methods, such as SPH (Smoothed Particle Hydrodynamics) method (Gingold and Monaghan, 1977; Lucy, 1977), and MPS (Moving Particle Semi-implicit) one (Koshizuka and Oka, 1996; Khayyer and Gotoh, 2009; Khayyer and Gotoh, 2016) to simulate effectively such complicated flow problems.
Recently, the physics-based computer simulations on the GPU (Graphics Processing Units) (Harada et al., 2008; Green, 2010; Hori et al., 2011) have increasingly become an important strategy for solving efficiently various problems, such as fluid dynamics, solid dynamics, and so forth. In our previous work, we have presented a GPU-based MPS scheme using logarithmic weighting function for solving effectively 2D/3D problems of incompressible fluid flow (Kakuda et al., 2012).
Accidents of ships and offshore structures occur for various reasons. However, the flooding process always occurs before ships and offshore structures finally sinks, and this process takes a relatively long time due to the design characteristics of ships and offshore structures. Therefore, accurate analysis and understanding of these flooding process are necessary to minimize the damage of human life on ships and offshore structures. For analyzing this process, in this study, Position Based Dynamics (PBD) which is one of the mesh-free particle methods was applied. In this study, the method was applied to various examples of flooding case, and the availability of the method was evaluated.
Using PBD, the dynamic effects on ships and offshore structures that are difficult to be analyzed in the quasi-static method can be considered. In addition, most of the problems that arise when analyzing with existing CFDs (Computational Fluid Dynamics) can be solved.
Traditionally, the design rules for ships and offshore structures were judged whether they are satisfied with stability only by the final state of ships and offshore structures after damage. Therefore, there was a lack of understanding of the flooding process and time.
The flooding process of a damaged ships and offshore structures is a complex process which can involve various phenomena, such as collapsing of non-watertight structures and compression of air. In general, the flooding process that follows the creation of the damage opening can be divided into three main phases (IMO SLF46/INF.3, 2003). After a damage, there is followed by a phase of progressive flooding as the water floods to undamaged compartments through the internal openings such as doors and pipes (Ruponen, 2007). If the ship does not collapse or sink during these phases, a steady final state is eventually achieved. A schematic representation of these phases is shown in Fig. 1.
Until now, most related study focused on progressive flooding and assumed that the environmental effects including wind and wave is small. Therefore, they performed quasi-static analysis to simulate the flooding process. The quasi-static analysis is relatively fast and easy to generate the object model such as damaged ship and opening. However, as mentioned previously, damaged ships and offshore structures are subjected to dynamic effects by various environmental loads such as wave forces, current effects, wind forces and so on. In addition, the behavior of damaged floater in the transient phase may be greater than the progressive flooding phase for various reasons. Therefore, in this study, one of the mesh-free particle methods which is very useful to reflect dynamic effects was used. Mesh-free particle method does not need to consider mesh for fluid analysis, and there is an advantage that fluid analysis can be easily performed even if the size and position of domain change based on the behavior of ships and offshore structures. In addition, it is easy to apply the parallel computing method that has begun to be utilized in recent years to MPM, and it is much faster than the conventional grid-based fluid analysis methods. Therefore, the flooding analysis of ships and offshore structures based on MPM was proposed. Among various method in MPMs, flooding analysis was performed using PBD which have high speed and efficiency. PBD is a method developed in the field of computer graphics, and its accuracy is still being verified. It is very useful for computing speed and large amount of computation, but the accuracy can be lower than the conventional method. Because in PBD, speed and controllability are the important factors and all that is required in terms of accuracy is visual plausibility (Bender et al., 2014).
This contribution addresses the applicability of an efficient lattice Boltzmann-based single-phase free-surface model for the simulation of wave impact on the side walls of 2-D containers. The computational efficiency of the method is known to allow for very short turnaround times, but wave impact simulations have not been investigated in detail yet. Results for a selected wave impact case are discussed, the convergence behavior in space and time is analyzed, and limitations of the single-phase free-surface model are revealed. The results show that lattice Boltzmann method (LBM)-based single-phase free-surface models are a viable tool for predicting the impact wave behavior, but the quality of the pressure signal is limited, because of the absence of air in the simulations.
The efficient numerical simulation of violent tank sloshing and wave impact is important to many different fields of engineering. Besides the numerical accuracy of the employed solvers, the computational efficiency and the time to solution are of interest as well, as even two-dimensional simulations of tank sloshing require a substantial amount of computational time. In this context, a very efficient numerical methodology based on the lattice Boltzmann method (LBM) is assessed in this paper. The LBM is an alternative to conventional methods on the basis of the Navier–Stokes equations that offers solver-specific advantages in terms of data locality and parallel computing. The LBM usually operates on a finite difference grid, is explicit in time, and requires only next neighbor interaction. It is very suitable for implementation on graphics processing units (GPUs) and other high-performance computing (HPC) hardware. Recently published LB results comprise laminar and turbulent bulk flows, multiphase flows, and free-surface flow applications. For all applications, a comparably high computational performance on both CPU- and GPU-based parallel architectures is reported.
In the scope of this contribution, the applicability of the LBM to tank sloshing and wave impact is analyzed. Emphasis is given to the actual result accuracy, times to solution, and potential problems of the free-surface model. First, a short description of the LBM for bulk flows and the employed LBM free-surface model is given before addressing the violent tank sloshing case. Finally, conclusions are drawn and future perspectives are discussed.