Peter Chow

GPU accelerated CAE using open solvers and the cloud

After more than five years since GPUs were first used as accelerators for general scientific computations, the field of General Purpose GPU computing or GPGPU has finally reached mainstream. Developers have now access to a mature hardware and software ecosystem. On the software side, several major open-source packages now support GPU acceleration while on the hardware side cloud-based solutions provide a simple way to access powerful machines with the latest GPUs at low cost. In this context, we look at the GPU acceleration of CAE, with a focus on the matrix solvers. We compare the performance that can be achieved using the open-source solver package PETSc ran on GPU-enabled Amazon EC2 hardware with that of an optimized legacy FEM code ran on a last generation 12-core blade server. Our results show that, although good performance can be achieved, some development is still needed to achieve peak performance.

A stable FDTD subgridding method for both spatial and temporal spaces

The proposed FDTD subgridding method is stable in both spatial and temporal spaces. Late-time instability has been resolved by separating the space and time interpolation interfaces from conventional collocated to same position. Numerical experiments conducted in 2D of 10 million time steps have no instability problem after a minimum gap distance for various refinement factors. Compare to existing subgridding methods the new method is not restricted to a single refinement factor and need no special treatment at material interfaces. This should open the FDTD subgridding method to more applications.

Collaborating components in mesh-based electronic packaging

From the model geometry creation to the model analysis, the stages in between such as mesh generation are the most manpower intensive phase in a mesh-based computational mechanics simulation process. On the other hand the model analysis is the most computing intensive phase. Advanced computational hardware and software have significantly reduced the computing time - and more importantly the trend is downward. With the kind of models envisaged coming, which are larger, more complex in geometry and modelling, and multiphysics, there is no clear trend that the manpower intensive phase is to decrease significantly in time - in the present way of operation it is more likely to increase with model complexity. In this paper we address this dilemma in collaborating components for models in electronic packaging application.

Software design for decoupled parallel meshing of CAD models

The creation of Finite Element (FE) meshes is one of the most time-consuming steps in FE analysis. While the exponential increase in computational power, following Moores law, has gradually reduced the time spent in the FE solver, this has not generally been the case for FE mesh creation software. There are two main reason why this has been the case: most FE meshers are still serial and human intervention is generally required. In this paper we present the design of a system that tackles both these issues. More specifically, this paper proposes a system that, in combination with an unmodified off-the-shelf serial meshing program and an off-the-shelf CAD kernel, results in a fast and scalable tool capable of meshing complex CAD models, such as the ones used in industry, with reduced user intervention. To achieve scalability, our system uses two levels of parallelism: assembly level parallelism - across the multiple parts found in an assembly-type CAD model, and part level parallelism - obtained by partitioning individual CAD solids in multiple sections at the CAD level. We show preliminary results for the parallel meshing of a complex laptop model via which we highlight both some of the achieved benefits and the main challenges that need to be addressed in order to obtain good scalability.

Electronic Packaging and Reduction in Modelling Time Using Domain Decomposition

The domain decomposition method is directed to electronic packaging simulation in this article. The objective is to address the entire simulation process chain, to alleviate user interactions where they are heavy to mechanization by component approach to streamline the model simulation process.

Development of CAD-to-CAE Model Preparation Technology

This paper explains and demonstrates how to reduce time for preparation of 3-dimensional (3D) geometrical Computer-Aided-Engineering (CAE) model from 3D Computer-Aided-Design (CAD) data. In generally, CAE model preparation is labor intensive and takes long time. Main part of preparation work is simplification of 3D-CAD data to decrease mesh scale and without impacting the solution accuracy. The purpose of this study is to create automatic CAE model preparation technology for reduction of preparation time. In this study, automatic model preparation method is developed by using of geometrical and topological information of 3D-CAD data. Benchmark test is performed to proof the efficiency of the method.

A stable three-dimensional subgridding method for both spatial and temporal spaces

In this paper we detailed a new three-dimensional Subgridding or Multigrid-FDTD method based on our recently proposed method in two-dimensions. First a 3D spatial subgridding method based on Monks two-dimensional method is described. Then by using conservation principle we showed the spatial subgridding method is stable, as well as the temporal subgridding method used. Finally, by separating the temporal and spatial perimeter interfaces for subgridding to different positions — the “temporal-spatial” coupled problem is decoupled into two separate independent problems. This decoupling makes the new 3D subgridding method naturally stable for electromagnetic wave propagation problems in both spatial and temporal spaces. Numerical results obtained show no late-time instability after ten million time steps.

A multilevel-multigrid approach to multiscale electromagnetic simulation

The time-dependent Maxwells equations are solved for mobile device applications using a multilevel-multigrid finite-difference time-domain (FDTD) method. For three-dimensional models that simulate system level details of mobile devices, the smallest features are in the nanometre (10−−9 m) range, leading to a time-step size in the attosecond (10−−18 s) range. The feature sizes of mobile devices are in the centimetre (10−−2 m) range, while for health and safety studies that include human models features are in the metre range.

A novel HE-coupling for explicit multigrid-FDTD

In this paper we put forward a novel coupling method that directly ties both the field variables, electric (E) and magnetic (H), in the multigrid-FDTD algorithm. Conventional leapfrog multigrid-FDTD or subgridding only directly couples one field variable between the grids, either the E or the H field, and the one that is not coupled is indirectly linked with the first by the FDTD equations. The FDTD method staggers the E and H fields. This means a strong numerical coupling for the directly connected field but only a weak coupling for the second field for multigrid-FDTD. There is nothing to prevent the weakly connected field from disengaging from each other in different grids, and if it does, it leads to instability. This can happen because there is nothing explicit that ties them together. Finally, a new and coherent approximation has been derived using the divergence and Greens theorems for the spatial derivative terms in the FDTD equations to enhance stronger coupling in multigrid-FDTD.

A Software Strategy towards Putting Domain Decomposition at the Centre of a Mesh-Based Simulation Process

In the mesh-based computer aided engineering sphere the most manpower intensive stages in the simulation process chain are the model creation and mesh generation. Together, they can account for 60 to 90 per cent of the loial modelling lime for complex industrial 3D models. In component meshing and gluing (CMG) we take a novel approach, using domain decomposition technology, to try to reduce the total modelling time further within the existing process without incurring significant re-engineering costs. The thinking is to offload a sizeable portion of the workload performed by manpower in the creation and meshing stages to the downstream stage, the analysis, performed by computing horsepower. In this paper we describe the concept and software strategy for such an approach.