arXiv subjects

Freddie Witherden

Publications and source records attributed to Freddie Witherden.

14 recordsLinked to original sources

arXiv2025

Analyzing the Impact of Augmented Reality Head-Mounted Displays on Workers' Safety and Situational Awareness in Hazardous Industrial Settings

Augmented Reality Head-Mounted Displays (AR-HMDs) have proven effective to assist workers. However, they may degrade their Safety and Situational Awareness (SSA), particularly in complex and hazardous industrial settings. This paper analyzes, objectively and subjectively, the effects of AR-HMDs' on workers' SSA in a simulated hazardous industrial environment. Our evaluation was comprised of sixty participants performing various tasks in a simulated cargo ship room while receiving remote guidance through one of three devices: two off-the-shelf AR-HMDs (Trimble XR10 with HoloLens 2, RealWear Navigator 520), and a smartphone (Google Pixel 6). Several sensors were installed throughout the room to obtain quantitative measures of the participants' safe execution of the tasks, such as the frequency in which they hit the objects in the room or stepped over simulated holes or oil spills. The results reported that the Trimble XR10 led to statistically highest head-knockers and knee-knocker incidents compared to the Navigator 520 and the Pixel 6. Furthermore, the Trimble XR10 also led to significantly higher difficulties to cross hatch doors, lower perceived safety, comfort, perceived performance, and usability. Overall, participants wearing AR-HMDs failed to perceive more hazards, meaning that safety-preserving capabilities must be developed for AR-HMDs before introducing them into industrial hazardous settings confidently.

cs.HC↗

arXiv2023

Cache Blocking for Flux Reconstruction: Extension to Navier-Stokes Equations and Anti-aliasing

In this article, cache blocking is implemented for the Navier Stokes equations with anti-aliasing support on mixed grids in PyFR for CPUs. In particular, cache blocking is used as an alternative to kernel fusion to eliminate unnecessary data movements between kernels at the main memory level. Specifically, kernels that exchange data are grouped together, and these groups are then executed on small sub-regions of the domain that fit in per-core private data cache. Additionally, cache blocking is also used to efficiently implement a tensor product factorisation of the interpolation operators associated with anti-aliasing. By using cache blocking, the intermediate results between application of the sparse factors are stored in per-core private data cache, and a significant amount of data movement from main memory is avoided. In order to assess the performance gains a theoretical model is developed, and the implementation is benchmarked using a compressible 3D Taylor-Green vortex test case on both hexahedral and prismatic grids, with third- and forth-order solution polynomials. The expected performance gains based on the theoretical model range from 1.99 to 2.62, and the speedups obtained in practice range from 1.67 to 3.67 compared to PyFR v1.11.0.

physics.flu-dyn↗

arXiv2023

An efficient GPU-based h-adaptation framework via linear trees for the flux reconstruction method

In this paper, we develop the first entirely graphic processing unit (GPU) based h-adaptive flux reconstruction (FR) method with linear trees. The adaptive solver fully operates on the GPU hardware, using a linear quadtree for two dimensional (2D) problems and a linear octree for three dimensional (3D) problems. We articulate how to efficiently perform tree construction, 2:1 balancing, connectivity query, and how to perform adaptation for the flux reconstruction method on the GPU hardware. As a proof of concept, we apply the adaptive flux reconstruction method to solve the inviscid isentropic vortex propagation problem on 2D and 3D meshes to demonstrate the efficiency of the developed adaptive FR method on a single GPU card. Depending on the computational domain size, acceleration of one or two orders of magnitude can be achieved compared to uniform meshing. The total computational cost of adaption, including tree manipulations, connectivity query and data transfer, compared to that of the numerical solver, is insignificant. It can be less than 2% of the total wall clock time for 3D problems even if we perform adaptation as frequent as every 10 time steps with an explicit 3-stage Runge--Kutta time integrator.

physics.flu-dyn↗

arXiv2022

Nonlinear p-multigrid preconditioner for implicit time integration of compressible Navier--Stokes equations

Within the framework of $ p $-adaptive flux reconstruction, we aim to construct efficient polynomial multigrid ($p$MG) preconditioners for implicit time integration of the Navier--Stokes equations using Jacobian-free Newton--Krylov (JFNK) methods. We hypothesise that in pseudo transient continuation (PTC), as the residual drops, the frequency of error modes that dictates the convergence rate gets higher and higher. We apply nonlinear $p$MG solvers to stiff steady problems at low Mach number ($\mathrm{Ma}=10^{-3}$) to verify our hypothesis. It is demonstrated that once the residual drops by a few orders of magnitude, improved smoothing on intermediate $ p $-sublevels will not only maintain the stability of $ p $MG at large time steps but also improve the convergence rate. For the unsteady Navier--Stokes equations, we elaborate how to construct nonlinear preconditioners using pseudo transient continuation for the matrix-free generalized minimal residual (GMRES) method used in explicit first stage, singly diagonally implicit Runge--Kutta (ESDIRK) methods, and linearly implicit Rosenbrock--Wanner (ROW) methods. Given that at each time step the initial guess in the nonlinear solver is not distant from the converged solution, we recommend a two-level $p\{p_0\text{-}p_0/2\} $ or even $ p\{p_0\text{-}(p_0-1)\} $ $p$-hierarchy for optimal efficiency with a matrix-based smoother on the coarser level based on our hypothesis. It is demonstrated that insufficient smoothing on intermediate $p$-sublevels will deteriorate the performance of $p$MG preconditioner greatly. (See full abstract in the paper.)

math.NA↗

arXiv2021

Artificial Compressibility Approaches in Flux Reconstruction for Incompressible Viscous Flow Simulations

Several competing artificial compressibility methods for the incompressible flow equations are examined using the high-order flux reconstruction method. The established artificial compressibility method (ACM) of \citet{Chorin1967} is compared to the alternative entropically damped (EDAC) method of \citet{Clausen2013}, as well as an ACM formulation with hyperbolised diffusion. While the former requires the solution to be converged to a divergence free state at each physical time step through pseudo iterations, the latter can be applied explicitly. We examine the sensitivity of both methods to the parameterisation for a series of test cases over a range of Reynolds numbers. As the compressibility is reduced, EDAC is found to give linear improvements in divergence whereas ACM yields diminishing returns. For the Taylor--Green vortex, EDAC is found to perform well; however on the more challenging circular cylinder at $Re=3900$, EDAC gives rise to early transition of the free shear-layer and over-production of the turbulence kinetic energy. This is attributed to the spatial pressure fluctuations of the method. Similar behaviour is observed for an aerofoil at $Re=60,000$ with an attached transitional boundary layer. It is concluded that hyperbolic diffusion of ACM can be beneficial but at the cost of case setup time, and EDAC can be an efficient method for incompressible flow. However, care must be taken as pressure fluctuations can have a significant impact on physics and the remedy causes the governing equation to become overly stiff.

physics.flu-dyn↗

arXiv2021

Hyperbolic Diffusion in Flux Reconstruction: Optimisation through Kernel Fusion within Tensor-Product Elements

Novel methods are presented in this initial study for the fusion of GPU kernels in the artificial compressibility method (ACM), using tensor product elements with constant Jacobians and flux reconstruction. This is made possible through the hyperbolisation of the diffusion terms, which eliminates the expensive algorithmic steps needed to form the viscous stresses. Two fusion approaches are presented, which offer differing levels of parallelism. This is found to be necessary for the change in workload as the order of accuracy of the elements is increased. Several further optimisations of these approaches are demonstrated, including a generation time memory manager which maximises resource usage. The fused kernels are able to achieve 3-4 times speedup, which compares favourably with a theoretical maximum speedup of 4. In three dimensional test cases, the generated fused kernels are found to reduce total runtime by ${\sim}25\%$, and, when compared to the standard ACM formulation, simulations demonstrate that a speedup of $2.3$ times can be achieved.

cs.MS↗

arXiv2021

Turbulence closure modeling with data-driven techniques: Investigation of generalizable deep neural networks

Generalizability of machine-learning (ML) based turbulence closures to accurately predict unseen practical flows remains an important challenge. At the Reynolds-averaged Navier-Stokes (RANS) level, NN-based turbulence closure modeling is rendered difficult due to two important reasons: inherent complexity of the constitutive relation arising from flow-dependent non-linearity and bifurcations; and, inordinate difficulty in obtaining high-fidelity data covering the entire parameter space of interest. In this context, the objective of the work is to investigate the approximation capabilities of standard moderate-sized fully-connected NNs. We seek to systematically investigate the effects of: (i) intrinsic complexity of the solution manifold; (ii) sampling procedure (interpolation vs. extrapolation) and (iii) optimization procedure. To overcome the data acquisition challenges, three proxy-physics turbulence surrogates of different degrees of complexity (yet significantly simpler than turbulence physics) are employed to generate the parameter-to-solution maps. Even for this simple proxy-physics system, it is demonstrated that feed-forward NNs require more degrees of freedom than the original proxy-physics model to accurately approximate the true model even when trained with data over the entire parameter space (interpolation). Additionally, if deep fully-connected NNs are trained with data only from part of the parameter space (extrapolation), their approximation capability reduces considerably and it is not straightforward to find an optimal architecture. Overall, the findings provide a realistic perspective on the utility of ML turbulence closures for practical applications and identify areas for improvement.

physics.flu-dyn↗

arXiv2021

Partially-Averaged Navier-Stokes Simulations of Turbulence Within a High-Order Flux Reconstruction Framework

High-order methods and hybrid turbulence models have independently shown promise as means of decreasing the computational cost of scale-resolving simulations. The objective of this work is to develop the combination of these methods and analyze the effects of high-order discretizations on hybrid turbulence models, particularly with respect the optimal model parameters and the relative accuracy benefits compared to approaches such as under-resolved direct numerical simulation (URDNS). We employ the Partially-Averaged Navier-Stokes (PANS) approach using the flux reconstruction scheme on the flow around a periodic hill and the wake flow of a circular cylinder at a Reynolds number of 3900, the latter of which we provide direct numerical simulation results and novel statistical analysis. By increasing the order of the discretization while fixing the total degrees of freedom, it was observed that larger improvements in the prediction of the statistics and flow physics were generally seen with PANS than URDNS. Furthermore, less sensitivity to the resolution-control parameter was observed with a high-order discretization, indicating that high-order discretizations may be an effective approach for increasing the accuracy and reliability of hybrid turbulence models for scale-resolving simulations without a significant increase in computational effort.

physics.flu-dyn↗

arXiv2020

Fourier Spectrum Discrepancies in Deep Network Generated Images

Advancements in deep generative models such as generative adversarial networks and variational autoencoders have resulted in the ability to generate realistic images that are visually indistinguishable from real images, which raises concerns about their potential malicious usage. In this paper, we present an analysis of the high-frequency Fourier modes of real and deep network generated images and show that deep network generated images share an observable, systematic shortcoming in replicating the attributes of these high-frequency modes. Using this, we propose a detection method based on the frequency spectrum of the images which is able to achieve an accuracy of up to 99.2% in classifying real and deep network generated images from various GAN and VAE architectures on a dataset of 5000 images with as few as 8 training examples. Furthermore, we show the impact of image transformations such as compression, cropping, and resolution reduction on the classification accuracy and suggest a method for modifying the high-frequency attributes of deep network generated images to mimic real images.

eess.IV↗

arXiv2020

Accuracy, Stability, and Performance Comparison between the Spectral Difference and Flux Reconstruction Schemes

We report the development of a discontinuous spectral element flow solver that includes the implementation of both spectral difference and flux reconstruction formulations. With this high order framework, we have constructed a foundation upon which to provide a fair and accurate assessment of these two schemes in terms of accuracy, stability, and performance with special attention to the true spectral difference scheme and the modified spectral difference scheme recovered via the flux reconstruction formulation. Building on previous analysis of the spectral difference and flux reconstruction schemes, we provide a novel nonlinear stability analysis of the spectral difference scheme. Through various numerical experiments, we demonstrate the additional stability afforded by the true, baseline spectral difference scheme without explicit filtering or de-aliasing due to its inherent feature of staggered flux points. This arrangement leads to favorable suppression of aliasing errors and improves stability needed for under-resolved simulations of turbulent flows.

physics.comp-ph↗

arXiv2020

A New Family of Weighted One-Parameter Flux Reconstruction Schemes

The flux reconstruction (FR) approach offers a flexible framework for describing a range of high-order numerical schemes; including nodal discontinuous Galerkin and spectral difference schemes. This is accomplished through the use of so-called correction functions. In this study we employ a weighted Sobolev norm to define a new extended family of FR correction functions, the stability of which is affirmed through Fourier analysis. Several of the schemes within this family are found to exhibit reduced dissipation and dispersion overshoot. Moreover, many of the new schemes possess higher CFL limits whilst maintaining the expected rate of convergence. Numerical experiments with homogeneous linear convection and Burgers turbulence are undertaken, and the results observed to be in agreement with the theoretical findings.

math.NA↗

arXiv2020

Inline Vector Compression for Computational Physics

A novel inline data compression method is presented for single-precision vectors in three dimensions. The primary application of the method is for accelerating computational physics calculations where the throughput is bound by memory bandwidth. The scheme employs spherical polar coordinates, angle quantisation, and a bespoke floating-point representation of the magnitude to achieve a fixed compression ratio of 1.5. The anisotropy of this method is considered, along with companding and fractional splitting techniques to improve the efficiency of the representation. We evaluate the scheme numerically within the context of high-order computational fluid dynamics. For both the isentropic convecting vortex and the Taylor--Green vortex test cases, the results are found to be comparable to those without compression. Performance is evaluated for a vector addition kernel on an NVIDIA Titan V GPU; it is demonstrated that a speedup of 1.5 can be achieved.

cs.CE↗

arXiv2018

High-Order Computational Fluid Dynamics Simulations of a Spinning Golf Ball

This paper presents the first high-order computational fluid dynamics (CFD) simulations of static and spinning golf balls at realistic flow conditions. The present results are shown to capture the complex fluid dynamics inside the dimples which lead to drag reduction versus a smooth sphere, and compare well to previous experimental and computational studies. The high--order Flux Reconstruction method has been paired with the Artificial Boundary overset method to enable simplified mesh generation and grid motion. The compressible Navier--Stokes equations are modeled using a scale--resolving Large Eddy Simulation (LES) approach with no sub--grid models. The codes implementing these methods have been implemented for NVIDIA Graphical Processing Units (GPUs), enabling large speedups over traditional computer hardware. The new method allows for the simulation of golf balls, and other objects at similar moderate Reynolds numbers, to be simulated in a matter of days on large computing clusters. The use of CFD for the design of objects such as golf balls and other sports balls is now within reach.

physics.flu-dyn↗

arXiv2016

Using the pyMIC Offload Module in PyFR

PyFR is an open-source high-order accurate computational fluid dynamics solver for unstructured grids. It is designed to efficiently solve the compressible Navier-Stokes equations on a range of hardware platforms, including GPUs and CPUs. In this paper we will describe how the Python Offload Infrastructure for the Intel Many Integrated Core Architecture (pyMIC) was used to enable PyFR to run with near native performance on the Intel Xeon Phi coprocessor. We will introduce the architecture of both pyMIC and PyFR and present a variety of examples showcasing the capabilities of pyMIC. Further, we will also compare the contrast pyMIC to other approaches including native execution and OpenCL. The process of adding support for pyMIC into PyFR will be described in detail. Benchmark results show that for a standard cylinder flow problem PyFR with pyMIC is able achieve 240 GFLOP/s of sustained double precision floating point performance; for a 1.85 times improvement over PyFR with C/OpenMP on a 12 core Intel Xeon E5-2697 v2 CPU.

cs.MS↗