Anbieter zum Thema

ALPHA-Numerics GmbH

Electric vehicles need solutions to remove excess heat

In automotive applications, the situation can be even more challenging to the designer. Hybrid and electric vehicles need high-density power electronics but the difficulty to remove excess heat can reduce reliability and possibly even lead to expensive, high-profile vehicle recalls.

An understanding of the distribution of heat through a system can also be used to tune a design. There are situations where increasing the ambient temperature of some components on a board, such as diodes, can reduce electrical losses, while others suffer badly in terms of reliability and performance if not kept as cool as possible. Highly accurate, fast thermal modelling allows engineering teams to assess intra-system thermal performance accurately and quickly.

Thermal simulation is a specialisation of the technology developed for computational fluid dynamics (CFD) to model the flow of fluids through and around objects using the Navier-Stokes flow equations. A model is created using intelligent modelling objects, CAD models and PCB layout information.

To be made ready for simulation, the system’s construction needs to be represented as a mesh of grid cells. There are many different methods of generating a suitable grid to be solved. The quality of grid is crucial to achieving an accurate solution.

There is a trade-off between compute performance and detail in the grid. A grid that is too finely detailed will not only take much longer to calculate but may also not even fit into the computer’s memory.

But, if the grid is simplified to uses a more coarse-grained structure, there can be a significant impact on the accuracy of the simulation in places where finer detail should have been employed. In practice, thermal simulation engineers will alter the resolution of the grid to focus on those parts of the design that are thought to be thermally significant.

But this fine-tuning of grid resolution at a local level has traditionally required manual intervention, which is a time consuming and complex process. The key to effective simulation is to provide a more effective way of optimising the grid to provide both performance and accuracy where it is needed.

The latest release of 6SigmaET incorporates a multi-level unstructured staggered grid solver that provides a novel way to efficiently solve the key Navier-Stokes equations and, as importantly, relieves the user from the burden of having to manually optimise the grid.

Starting with a fine grid over the whole model, a hierarchy of increasingly coarse Cartesian grids is constructed and the solver selects from these only those cells necessary to resolve the geometry. This ensures that large areas that do not call for highly detailed grids are treated at a suitable level of resolution, but still allow the tool to select a finer grid for smaller objects.

The approach used in Release 9 avoids many problems associated with alternative unstructured techniques and provides ground-breaking performance. For example, a design that previously demanded nearly one billion grid cells has been simplified to a model that employs just six million cells, providing a solution in a matter of minutes. Tenfold speed-ups have been demonstrated on many designs.

Crucially, although it simplifies the grid, the solver’s approach does not lose accuracy vital to modelling electronics systems. Improvements in the way grids are generated are making it much easier to build accurate simulations of electronic systems.

(ID:42994202)