Although computational simulation is extensively employed in industry, its wider use is limited by the complexity of the geometric models involved. This limitation is due to the excessive number of human hours, ranging from days to months, required to transfer information from a computer aided design (CAD) model to a computer aided engineering (CAE) model suitable for simulation. CAD models frequently involve a level of detail much greater than that required to perform a computational simulation with a CAE system. The preparation of CAD models for simulation, including mesh generation, is still a challenging bottleneck that needs to be resolved to enable realisation of the full potential of simulation tools in industry. This challenge is also a crucial factor delaying the industrial uptake of, the often computationally superior, high-order methods.
Current research is focused on the development of algorithms for de-featuring complex CAD models. A major drawback of this process is the requirement for human expertise and manual interaction with CAD systems and geometry cleaning tools. Although engineers are aided by the semi-automatic tools that are included in many existing commercial mesh generation packages, such as COMSOL, ANSYS, CATIA, SolidWorks, Patran, MSC, CADfix, ESI Visual Environment, de-featuring cannot be fully automatised. In addition, it is usually not possible to know, a priori, the effect of de-featuring on the results of a simulation because this process depends upon the physical problem and the level of approximation required.
At the heart of the problem is the traditional hierarchical paradigm implemented in many commercial mesh generators. The ultimate goal of this project is to develop a new computational environment that includes a feature-independent mesh generation paradigm and plug--and--play libraries to enable direct integration of the meshes into existing commercial and research solvers. The proposed approach is disruptive, as it proposes the development of unconventional computational approaches, not only at the stage of generating suitable meshes for computational simulations but also requires the incorporation of new plug-and-play libraries into existing solvers. The libraries will be delivered as part of this project and it will follow the rationale used in commercial software where the user can select a different type of element depending on the demands of a particular simulation.
The advantage of the proposed mesh generation technique is not restricted to removing the bottleneck that has been highlighted by many industries that routinely use computational engineering in their design cycles. In addition, the new meshes will completely remove the uncertainty introduced by de-featuring CAD models. Instead of relying on the opinion of experts, to decide which features might not be relevant in a simulation, the CAD model will not be altered, leading to higher fidelity simulations and more confidence in the results.
The proposed research is timely, tackling a problem that has been highlighted in the last three years by independent agencies (e.g. NASA), international associations dedicated to computer modelling (e.g. NAFEMS) and the private sector (e.g. Pointwise Inc.). Since the mid 1990s the research has focused on the development of tools for faster de-featuring. The fact that this issue has not been resolved in over two decades, suggests that the radical new approach proposed here, pursuing an orthogonal research direction, in which no de-featuring is needed, can lead to a breakthrough.
|