Parallel Numerical Model Research Articles

A Hybrid Parallel Numerical Model for Wave-Induced Free-Surface Flow

An advanced numerical model is presented for the simulation of wave-induced free-surface flow, utilizing an efficient hybrid parallel implementation. The model is based on the solution of the Navier–Stokes equations using large-eddy simulation of large-scale coastal free-surface flows. The three-dimensional immersed boundary method was used for the enforcement of the no-slip boundary condition on the bed surface. The water-air interface was tracked using the level-set method. The numerical model was effectively validated against laboratory measurements involving wave propagation over a flatbed with an elliptical shoal, whose presence induces combined wave refraction and diffraction phenomena. The parallel implementation of the model enabled the efficient simulation of depth-resolved, wave-induced, three-dimensional, free-surface flow; the model parallel efficiency and strong scaling are quantitatively demonstrated.

Numerical Simulations of 2D Hydraulic Jumps by a Parallel SPH Model

Hydraulic jumps are a rapid transition from supercritical to subcritical flow and generally occur in rivers or spillways. Owing to the high energy dissipation rate, hydraulic jumps are widely applied as energy dissipators in hydraulic projects. To achieve efficient and accurate simulations of 2D hydraulic jumps in open channels, a parallel Weakly Compressible Smoothed Particle Hydrodynamics model (WCSPH) with Shepard Density filter was established in this study. The acceleration of the model was obtained by OpenMP to reduce execution time. To further reduce execution time, a suitable and efficient scheduling strategy was selected for the parallel numerical model by comparing parallel speed-ups under different scheduling strategies in OpenMP. Following this, two test cases of uniform flow in open channels and hydraulic jumps with different inflow conditions were investigated to validate the model. The comparison of the water depth and velocity fields between the numerical results and the analytical solution generally showed good agreement, although there was a minor discrepancy in conjugate water depths. The numerical results showed free surface undulation with decreasing amplitude, which is more consistent with physical reality, with a low inflow Froude number. Simultaneously, the Shepard filter was able to smooth the pressure fields of the hydraulic jumps with a high inflow Froude number. Moreover, the parallel speed-up was generally able to reach theoretical maximum acceleration by analyzing the performance of the model according to different particle numbers.

Discrete-Event Simulation Thermal Model for Extrusion-Based Additive Manufacturing of PLA and ABS.

The material properties of thermoplastic polymer parts manufactured by the extrusion-based additive manufacturing process are highly dependent on the thermal history. Different numerical models have been proposed to simulate the thermal history of a 3D-printed part. However, they are limited due to limited geometric applicability; low accuracy; or high computational demand. Can the time–temperature history of a 3D-printed part be simulated by a computationally less demanding, fast numerical model without losing accuracy? This paper describes the numerical implementation of a simplified discrete-event simulation model that offers accuracy comparable to a finite element model but is faster by two orders of magnitude. Two polymer systems with distinct thermal properties were selected to highlight differences in the simulation of the orthotropic response and the temperature-dependent material properties. The time–temperature histories from the numerical model were compared to the time–temperature histories from a conventional finite element model and were found to match closely. The proposed highly parallel numerical model was approximately 300–500 times faster in simulating thermal history compared to the conventional finite element model. The model would enable designers to compare the effects of several printing parameters for specific 3D-printed parts and select the most suitable parameters for the part.

Parallel Geothermal Numerical Model with Fractures and Multi-Branch Wells

To answer the need for an efficient and robust geothermal simulation tool going beyond existing code capabilities in terms of geological and physical complexity, we have started to develop a parallel geothermal simulator based on unstructured meshes. The model takes into account complex geology including fault and fracture networks acting as major heat and mass transfer corridors and complex physics coupling the mass and energy conservations to the thermodynamic equilibrium between the gas and liquid phases. The objective of this Cemracs project was to focus on well modeling which is a key missing ingredient in our current simulator in order to perform realistic geothermal studies both in terms of monitoring and in terms of history matching. The well is discretized by a set of edges of the mesh in order to represent efficiently slanted or multi-branch wells on unstructured meshes. The connection with the 3D matrix and the 2D fracture network at each node of the well is accounted for using Peaceman’s approach. The non-isothermal flow model inside the well is based on the usual single unknown approach assuming the hydrostatic and thermodynamical equilibrium inside the well. The parallelization of the well model is implemented in such a way that the assembly of the Jacobian at each Newton step and the computation of the pressure drops inside the well can be done locally on each process without MPI communications.

Hydrograph estimation at upstream ungauged sections on the Secchia River (Italy) by means of a parallel Bayesian inverse methodology

In this work, we present a reverse flow routing procedure, which allows estimating discharge hydrographs at upstream ungauged stations by means of information available at downstream monitored sites. The reverse routing problem is solved adopting a Bayesian Geostatistical Approach (BGA). In order to capture the complex hydrodynamic field typical of many real cases of rivers including large floodable areas, meanwhile overcoming the computational time limitations, we adopted as forward model a selfdeveloped 2D-SWE parallel numerical model (PARFLOOD) that allows achieving ratio of physical to computational time of about 500-1000. To exploit the computational capabilities of modern GPU cluster, a parallel procedure to estimate the Jacobian matrix required by the BGA approach has been implemented. The inflow hydrograph in a river reach with several meanders and floodplains has been estimated in “only” 13 hours using a HPC cluster with 10 P100 Nvidia GPUs.

Parallel numerical modeling of hybrid-dimensional compositional non-isothermal Darcy flows in fractured porous media

This paper introduces a new discrete fracture model accounting for non-isothermal compositional multiphase Darcy flows and complex networks of fractures with intersecting, immersed and non-immersed fractures. The so called hybrid-dimensional model using a 2D model in the fractures coupled with a 3D model in the matrix is first derived rigorously starting from the equi-dimensional matrix fracture model. Then, it is discretized using a fully implicit time integration combined with the Vertex Approximate Gradient (VAG) finite volume scheme which is adapted to polyhedral meshes and anisotropic heterogeneous media. The fully coupled systems are assembled and solved in parallel using the Single Program Multiple Data (SPMD) paradigm with one layer of ghost cells. This strategy allows for a local assembly of the discrete systems. An efficient preconditioner is implemented to solve the linear systems at each time step and each Newton type iteration of the simulation. The numerical efficiency of our approach is assessed on different meshes, fracture networks, and physical settings in terms of parallel scalability, nonlinear convergence and linear convergence.

Finite element analysis of switched reluctance motor with rotor position based control

This research presents a field-circuit coupled parallel finite element model of a switched reluctance motor embedded in a simple closed loop control system. The parallel numerical model is based on the Schur-complement method coupled with an iterative solver. The used control system is the rotor position based control, which is applied to the FEM model. The results and parallel performance of the voltage driven finite element model are compared with the results from the current driven model. Moreover, the results of the start-up of the loaded motor show why the model accuracy is important in the control loop.

An efficient parallel numerical modeling of bioheat transfer in realistic tissue structure

A fast and accurate numerical method of solving bioheat transfer problems is critical for many biomedical applications, such as efficient implementation of thermal therapy planning in a clinical setup, and evaluation of the thermal effects induced by specific energy absorption rate (SAR) from external electromagnetic field. This paper developed a parallel finite-difference method based on alternating direction explicit (ADE) scheme to solve three dimensional transient bioheat equation for the realistic tissue structure. An optimized processing pipeline for accurate presentation of the irregular boundary condition was established to considerably reduce the staircase errors induced by cubic voxel. The voxels order aligned along diagonal direction was designed to allow fast parallel strategy for the lower and upper matrix solution involved in ADE scheme for irregular tissue. The test cases including a real head model have indicated that the developed method can support large time step and cubic voxel approximation of irregular interface and boundary, and also produce significant parallel efficiency.

Simulation of the world ocean climate with a massively parallel numerical model

The INM-IO numerical World Ocean model is verified through the calculation of the model ocean climate. The numerical experiment was conducted for a period of 500 years following the CORE-I protocol. We analyze some basic elements of the large-scale ocean circulation and local and integral characteristics of the model solution. The model limitations and ways they are overcome are described. The results generally fit the level of leading models. This experiment is a necessary step preceding the transition to high-resolution diagnostic and prognostic calculations of the state of the World Ocean and its individual basins.

Improvement of the Wet and Dry Algorithm for an Inundation Problem using a Parallel Numerical Model

Min, B.; Park, K.; Youm, M., and Suh, K., 2014. Improvement of the wet and dry algorithm for an inundation problem using a parallel numerical model.Rip current flow is dependent on wave-induced set up and thus on wave conditions and wave dissipation. Waves will break over the bar if the ratio of wave height to water depth exceeds a certain value. This implies that rip currents are not only dependent on wave height but also on water level that might be modified by the tide. Many studies noticed rip-current in tidal environments is that tide acts mainly as a change in water level. In coastal regions, the flooding and ebbing of the tide and waves play important roles in the local ecosystem and coastal physical processes. The complex physical process of a flood and ebbing wave presents a nontrivial modeling challenge. For a closed water body, where evaporation, precipitation and ground absorption are not considered, the total water mass should be conserved during inundation or draining. In this study, the ADCIRC (Advanced CIRCulation Model) was configured for a 120 by 120 km, square-shaped model domain. A bowl-shaped water system, with a surface diameter of 60 km and a maximum depth of 9 m, was embedded in the center of the model domain. A mass conserving wetting and drying scheme, which is very simple and practical. The water mass removed from any wet cell is uniformly redistributed into the system to maintain the water mass balance. Since there is a uniform change in the system, the hydrodynamic balance between the pressure gradient and the flow fields in the system is maintained. The redistribution of the water mass and readjustment of the sea surface elevation in the system are performed by the model in subsequent steps. The advanced wetting and drying scheme shows a good performance. This study was conducted as part of the wave-tide-surge coupling model (ADCIRC-SWAN) development which system can be used for predict the rip-current system combined with tidal environment.

Pedestrian Timber Bridges: Experimental Investigation and Modelling

For a specific pedestrian timber link case study, the authors had the chance to collect measurements of the structural response for a period of a few days but with the chance to repeat the monitoring in different seasonal weather conditions. This paper provides a synthesis of the data collected up to now. A parallel numerical model was implemented accounting for different potential constitutive law of the timber components and different hypotheses on the supports behavior.

Performance improvement with parallel numerical model simulations in the field of urban water management

Numerical models are used to enhance the understanding of the behavior of real world systems. With increasing complexity of numerical models and their applications, there is a need of more computational power. State of the art processors contain many cores on one single chip. As such, new programming techniques are required if all these cores are to be utilized during model simulation runs. This manuscript reviews the runtime and speedup behavior of parallel model analysis software (e.g. Calimero and Achilles) applied to simulation tools for urban water management (e.g. CityDrain3, EPANET2, SWMM5, par-SWMM). The potential of using a parallel programming environment for ‘coordinating’ tasks of multiple runs of commonly used modeling software is analyzed. This is especially interesting as the modeling software itself can be implanted sequentially or parallel. Performance tests are performed on a set of real-world case studies. Additionally, a benchmark set of 2,280 virtual case studies is used to investigate performance improvement in relation to the size of the system. It was found that speedup depends on the system size and the time spent in critical code sections with increasing number of used cores. Applying parallelism only at the level of the model analysis software performs best.

Impact response of layered steel–alumina targets

The efficacy of a ceramic in protecting against penetration by high velocity projectiles depends not only on its hardness but also on its resistance to flow after comminution. Here we investigate experimentally the response of a model armor system comprising an alumina tile and two steel face sheets subject to impact by steel spherical projectiles. Complementary experiments are performed on the face sheet materials and the ceramic alone in order to gain insights into the inelastic responses of the constituent materials. A parallel numerical modeling study is performed of the system response with emphasis on the shape of the back face following impact. To this end, we employ the ceramic deformation model developed by Deshpande and Evans and modified here to account for dilatational softening following full comminution. Comparisons between model predictions and experimental measurements demonstrate the important role of granular flow. Preliminary parametric studies further suggest that additional effort is required to understand the transition in mechanical response of a ceramic as it transforms from a monolithic solid to a densely-packed granulated medium.

Construction of a parallel computational model by composition of computational facilities

We consider a general algorithm for constructing a parallel numerical model by composition of several computational objects. The only requirement for the applicability of the proposed algorithm is for the disturbance propagation speed to be finite. We present one possible approach to map a numerical model onto an object-oriented program model. Some results of numerical experiments on solving the heat conduction equation in linear and nonlinear cases are discussed.

Virtual ChuanDian — A parallel numerical modeling of Sichuan-Yunnan regional strong earthquake activities: Model construction and parallel simulation

Reasonable use of mathematical modeling and numerical simulation of regional earthquake activities has attracted more and more attentions in realistic seismicity studies in recent years. The inherent nature of their attendant multi-scale properties, from either the temporal or the spatial direction, necessitates the use of the cutting-edge massive parallel data processing techniques. Our study here is focused on the model construction and numerical simulation of the regional stress evolution after large earthquake sequences along the fault zones in Sichuan and Yunnan areas, using large-scale parallel finite element approximations with tens of millions of unstructured meshes on the distributed systems. The objective of our study is to provide a state-of-the-art model and a large-scale numerical simulation platform, on which estimation of the risk of large earthquake triggering sequences can be carried out in a middle or long-term time scale with realistic data and constrains inputted from field observation and geological survey.

Parallel Numerical Model for the Simulation of Gravity Wave’s Nonlinear Propagation

基于MPI 消息传递接口和区域分解的思想, 对跳点网格上的串行数值模式进行并行化处理, 建立了模拟重力波非线性传播的并行数值模式. 针对跳点网格的特点, 详细论述了跳点网格系统下区域分解和各子区域间的数据通信. 对小振幅重力波传播过程的模拟结果表明, 并行数值模式可以很好地模拟小振幅重力波的传播过程, 也能保持重力波传播过程中的能量守恒关系. 并行数值模式的并行效率可以达到0.65, 在理想情况下可以达到最大值1.0. 作为与串行数值模式的比较, 采用不同网格分辨率以保证计算时间相同, 分别用串行和并行数值模式模拟有限振幅重力波的非线性传播过程, 结果表明, 通过引入更多的进程参与计算, 并行数值模式可以有效地提高模式的分辨率, 模拟出重力波在发生翻转之后, 而在演化成湍流之前的这段时间内有Kelvin-Helmholtz billows 出现, 但是, 在相同的计算时间内, 串行模式的分辨率较低, 不能刻画这些精细的现象.

Parallel Numerical Model for the Simulation of Three Dimensional Gravity Wave's Propagation

从三维大气运动的基本控制方程出发, 基于MPI消息传递接口和区域分解的思想, 建立了模拟三维可压缩大气中重力波传播过程的并行数值模式. 在对垂直方向进行区域分解的基础上, 针对跳点网格的特点, 以垂直速度主格点线为子区域的下边界, 状态变量主格点线为子区域的上边界进行区域分解. 利用MPI消息传递接口来传递计算各子区域中变量时所要用到的相邻子区域上相应变量, 从而顺利地完成对整个计算域的并行计算. 根据线性重力波理论, 通过模拟小振幅重力波的传播过程对所建模式进行了验证. 结果表明, 模式可以很好地模拟小振幅重力波在三维空间中的传播过程, 模拟的重力波振幅随着传播高度的增加以指数形式增长, 对能量传播路径以及有效扰动位能和扰动动能的模拟结果均与理论预测的结果吻合很好. 另外, 随着进程个数的增加, 完成相同的计算量所用的计算时间也显著减少. 这表明本文建立的并行数值模式不但能够很好的模拟重力波的传播过程, 而且能够有效地节约计算时间.

Modelling transient heat and moisture transfer in unsaturated soil using a parallel computing approach

AbstractA parallel numerical model, employing a finite difference explicit scheme for the analysis of coupled heat and moisture transfer in unsaturated soil, is employed to simulate a laboratory experiment of heating of medium sand. The model, written in a two‐dimensional polar co‐ordinate formulation, is programmed in the concurrent language Occam and executed on a parallel computing network of transputers. Parallelization is adopted as a means of overcoming computing difficulties, which limited numerical solutions to those at steady state, to enable transient behaviour to be simulated.The parallel algorithm was found to be very efficient, enabling a full solution of transient behaviour to be obtained. An investigation of the ability of the model to accurately simulate the complex, interrelated coupled nature of both two‐dimensional transient and steady‐state behaviour yielded very good correlation between experimental and numerical results. It can therefore be concluded that overall the results obtained provide confidence in the validity of the approach proposed.

The Fluid-Compensated Cement Bond Log

Summary Simulations of cement bond logging (CBL) have shown that wellbore fluid effects can be segregated from sonic-signal response to changing cement strengths. Traditionally, the effects have been considered negligible and the CBL's have been interpreted as if water were in the wellbore. However, large variations in CBL's have become apparent with the increasing number of logs run in completion fluids, such as CaCl2, ZnBr2, and CaBr2. To study wellbore fluid effects, physical and numerical models were developed that simulated the wellbore geometry. Measurements were conducted in 5-, 7-, and 9⅝-in. casings for a range of wellbore fluid types and for both densities and viscosities. Parallel numerical modeling used similar parameters. Results show that bond-log amplitudes varied dramatically with the wellbore fluid acoustic impedance—i.e., there was a 70% increase in signal amplitudes for 11.5 lbm/gal (1370-kg/m3) CaCl2 over the signal amplitude in water. This led to the development of a fluid-compensated bond log that corrects the amplitude for acoustic impedance of various wellbore fluids, thereby making the measurements more directly related to the cement quality.

The physics of radiation fog: I – a field study

The results of a field study of the evolution of radiation fog on three nights are presented. A parallel numerical model (Brown and Roach 1976 – referred to hereafter as II) was also developed to test ideas suggested by the principal features observed during the field study. These were: (i) Periods of significant fog development appeared to occur when wind speeds dropped below 0.5-1ms−1. (ii) The liquid water content of the fog was a small fraction of the total condensed out by cooling. The balance of water appears to have been deposited on the ground. (iii) Radiative cooling (deduced from radiative flux divergence measurements) was in general greater than the actual cooling. (iv) As the surface became radiatively shielded by developing fog, the radiation inversion migrated to the fog top, accompanied by the establishment of a convective regime with a slight super-adiabatic lapse rate in the lower part of the fog. (v) Ammonium sulphate was a principal constituent of condensation nuclei. It is shown that these features are consistent with the suggestion that the development of radiation fog is primarily controlled by a balance between radiative cooling, which encourages fog, and turbulence, which inhibits it. Gravitational settling of fog droplets and soil heat flux also emerge as important factors. The role of cloud microphysics is not passive, but is less clearly defined as yet. The numerical model (II) reproduces most of the observed features, but not the stepwise growth or sharpness of the fog top. Some of the practical implications of this work for forecasting and for fog modification are briefly discussed.