Search

This work proposes a novel multi-output neural network for the prediction of aerodynamic coefficients of aerofoils in two dimensions and wings in three dimensions. Contrary to existing neural networks that are often designed to predict aerodynamic quantities of interest, the proposed network considers as output the pressure at a number of selected points on the aerodynamic shape. The proposed multi-output neural network is compared with other approaches found in the literature. Furthermore, a detailed comparison of the proposed neural network with the popular proper orthogonal decomposition method is presented. The numerical results, involving high dimensional problems with flow and geometric parameters, show the benefits of the proposed approach.

The Yee finite difference time domain (FDTD) algorithm is widely used in computational electromagnetics because of its simplicity, low computational costs and divergence free nature. The standard method uses a pair of staggered orthogonal cartesian meshes. However, accuracy losses result when it is used for modelling electromagnetic interactions with objects of arbitrary shape, because of the staircased representation of curved interfaces. For the solution of such problems, we generalise the approach and adopt an unstructured mesh FDTD method. This co-volume method is based upon the use of a Delaunay primal mesh and its high quality Voronoi dual. Computational efficiency is improved by employing a hybrid primal mesh, consisting of tetrahedral elements in the vicinity of curved interfaces and hexahedral elements elsewhere. Difficulties associated with ensuring the necessary quality of the generated meshes will be discussed. The power of the proposed solution approach is demonstrated by considering a range of scattering and/or transmission problems involving perfect electric conductors and isotropic lossy, anisotropic lossy and isotropic frequency dependent chiral materials.

This paper introduces developments in modified techniques for the generation of unstructured, non-uniform, dual orthogonal meshes which are suitable for use with co-volume solution schemes. Two new mesh generation techniques, a modified advancing front technique and an octree-Delaunay algorithm, are coupled with a mesh optimisation algorithm. When using a Delaunay–Voronoi dual, to construct mutually orthogonal meshes for co-volume schemes, it is essential to minimise the number of Delaunay elements which do not contain their Voronoi vertex. These new techniques provide an improvement over previous approaches, as they produce meshes in which the number of elements that do not contain their Voronoi vertex is reduced. In particular, it is found that the optimisation algorithm, which could be applied to any mesh cosmetics problem, is very effective, regardless of the quality of the initial mesh. This is illustrated by applying the proposed approach to a number of complex industrial aerospace geometries.

The standard Yee algorithm is widely used in computational electromagnetics because of its simplicity and divergence free nature. A generalization of the classical Yee scheme to 3D unstructured meshes is adopted, based on the use of a Delaunay primal mesh and its high quality Voronoi dual. This allows the problem of accuracy losses, which are normally associated with the use of the standard Yee scheme and a staircased representation of curved material interfaces, to be circumvented. The 3D dual mesh leapfrog-scheme which is presented has the ability to model both electric and magnetic anisotropic lossy materials. This approach enables the modelling of problems, of current practical interest, involving structured composites and metamaterials.

This paper outlines a new computational aerodynamic design optimisation algorithm using a novel method of parameterising a computational mesh using ‘control nodes’. The shape boundary movement as well as the mesh movement is coupled to the movement of user–defined control nodes via a Delaunay Graph Mapping technique. A Modified Cuckoo Search algorithm is employed for optimisation within the prescribed design space defined by the allowed range of control node displacement. A finite volume compressible Navier–Stokes solver is used for aerodynamic modelling to predict aerodynamic design ‘fitness’. The resulting coupled algorithm is applied to a range of test cases in two dimensions including aerofoil lift–drag ratio optimisation intake duct optimisation under subsonic, transonic and supersonic flow conditions. The discrete (mesh–based) optimisation approach presented is demonstrated to be effective in terms of its generalised applicability and intuitiveness.

The use of numerical simulation as a design tool in engineering is nowadays routinely practiced. The initial stage of the simulation process is the generation of an appropriate mesh that is designed to capture the required physics. Mesh generation algorithms require water tight geometries as a starting point. Many automatic software have been developed in the last decades that are capable of ensuring water tightness with a minimum user interventions. However, when dealing with complex objects that contains multi-scale features it is often necessary to manually remove small features which are in the ‘expert’ opinion of the analyst will not have a measurable effect on the solution of the physical system under consideration. The major drawback of this de-featuring stage is the requirement for manual interaction with CAD systems or geometry c eaning tools. Furthermore, the de-featuring has to be repeated for every physical system that is required to be considered in the design, i.e fluid dynamics, electromagnetics, acoustics, heat and radiation, structure mechanics, etc. In addition, the de- featuring is often dependent on the level of approximation considered for the simulation of the physical system, i.e Euler, Reynolds average or LES in the case of fluid dynamics. In the case of electromagnetics and acoustics simulations, the de-featuring is dependent on every frequency of the required range the designer has to consider. In this presentation we propose an automatic unstructured mesh generation technique using a variant of the advancing front method whereby mesh fronts coincident with the NURBS boundary match the exact geometric definition from CAD independently on the element size. The method allows the possibility to have elements much larger than small geometric features, reducing the overall number of elements required and removing the need for physic dependent de-featuring. The method used to extend the gene ation technique to high order meshes, based on the approach presented in [1], will also be presented. The presentation will highlight the use of NURBS-Enhanced Finite Element Method [2] to incorporate the resulting meshes within Finite Element codes. Finally, examples will be shown to demonstrate the effectiveness of the proposed method.

This paper deals with fluidized sand simulation in order to estimate the impact of sand particle motion on the BLOODHOUND SuperSonic Car (SSC) drag forces, such phenomenon is known as a spray drag effect. A gas-particle model is used to simulate the sand particles that rise from the ground because of the strong shockwave-desert surface interaction. A finite volume scheme is used to discretise the continuous model with a special treatment of the solid phase equations. An indefinitely differentiable and anisotropic limiter to reinforce the method stability and reduce any excessive smearing is applied. To estimate the area where sand particles are detached from the ground, a criterion based on pressure change is proposed. The model is first validated on ac urved 90 � bend test case with comparison to experimental results and then applied to the supersonic car.

A method for incorporating multi-resolution capabilities within pre-existing global 3-D spherical mantle convection codes is presented. The method, which we term "geometric multigrid refinement", is based upon the application of a multigrid solver on non-uniform, structured grids and allows for the incorporation of local high-resolution grids within global models. Validation tests demonstrate that the method is accurate and robust, with highly efficient solutions to large-scale non-uniform problems obtained. Significantly, the scheme is conceptually simple and straightforward to implement, negating the need to reformulate and restructure large sections of code. Consequently, although more advanced techniques are under development at the frontiers of mesh refinement and solver technology research, the technique presented is capable of extending the lifetime and applicability of pre-existing global mantle convection codes.

This paper investigates the efficiency of a high-order nodal discontinuous Galerkin method for the numerical solution of Maxwell’s equations using hybrid meshes. An unstructured triangular or tetrahedral mesh is used near curved boundaries and a structured Cartesian mesh is used to fill the remainder of the domain. A quadrature-free implementation is employed for the regular quadrilateral and hexahedral elements which, together with the reduction in the number of internal faces, leads to a reduction in the cpu time requirements. Numerical examples in two and three dimensions are used to illustrate the benefits of using hybrid meshes.

This paper details the computational design optimisation strategy employed to achieve an engineering solution to the problem of excessive supersonic lift at the rear of the BLOODHOUND SSC (SuperSonic Car) during its design. The optimisation problem is described first, followed by details of the computational fluid dynamics procedure employed to study the aerodynamic performance of the vehicle and the design optimisation strategy utilising Design of Experiments. The `optimised' design resulting from this study is presented in the final section and contrasted with the `unoptimised' baseline geometry. The final vehicle geometry presented in this paper is, at the time of writing, being built and is due to be tested in 2013 in an attempt to increase the World Land Speed Record from 763 mph to 1,000 mph.

A procedure for generating curved meshes, suitable for high-order finite element analysis, is described. The strategy adopted is based upon curving a generated initial mesh with planar edges and faces by using a linear elasticity analogy. The analogy employs boundary loads that ensure that nodes representing curved boundaries lie on the true surface. Several examples, in both two and three dimensions, illustrate the performance of the proposed approach, with the quality of the generated meshes being analysed in terms of a distortion measure. The examples chosen involve geometries of particular interest to the computational fluid dynamics community, including anisotropic meshes for complex three dimensional configurations.

Considering the use of commercial-off-the-shelf (COTS) devices in industry, a case study has been undertaken implementing the goal driven optimisation of a venturi-type pre-mixer. This COTS device is used to supply the correct ratio of a specific fuel and oxidiser to a pre-mix burner. Providing the correct quantity of excess air results in a reduction of the NOx emissions; and this, in addition to the requirement for complete mixing of the constituents has been used to define the multi-objective optimisation of the device. A Computational Fluid Dynamics (CFD) simulation has been carried out using Fluent v14.5; verified by experimental data. To find an optimised design of the pre-mixer, four geometrical dimensions of its current design have been considered. An assessment of varying sampling and interpolation techniques has been made, in order to establish the most appropriate for use with this specific problem. The evolutionary-based NSGA-II and deterministic Nelder Mead optimisation algorithms have been used and their results compared to verify the optimisation outcome. Based on these findings suggested customisation of the COTS pre-mixer's design could result in improvements in mixing performance, in the region of 30 %.