Engineering Computational Technology
Chapter 4 C.G. Armstrong+, D.J. Monaghan+, M.A. Price+, H. Ou+ and J. Lamont*
+School of Mechanical & Manufacturing Engineering, Queen's University, Belfast, Northern Ireland *Transcendata Europe plc, Cambridge, England Keywords: CAE, CAE, feature suppression, dimensional reduction, dimensional addition, idealisation.
There now exists a new and growing body of knowledge associated with the use of computer based methods and solutions are increasingly tending towards integration of design processes and applications. The integration of CAE simulation models into the design process is a key factor to enabling enhanced product development. There has been substantial progress in CAx technologies that facilitate the integration of design and simulation but many gaps in functionality remain. An attempt to provide a coherent description of the future requirements is presented. Multidimensional geometric modelling with a consistent treatment of line, surface and solid models is an essential requirement for simulation modelling. Any geometric representation of an object may be divided into a system of rational subregions. However for the specification of analysis attributes, these cells will not generally correspond with the conventional B-Rep decomposition of the object, so a more general scheme for creating additional structure is required. Different analysis types, such as an FE analysis of a structure and a CFD analysis of the space surrounding it, could be specified by different cells of the same overall space. Cellular modelling is regarded as a key enabling technology for the next generation of CAE systems. Preliminary design investigations are routinely carried out on simple conceptual aerospace, automotive and construction models that have little or no linkage to the subsequent detailed designs. However, most of the information required to generate an initial 3D model is implicit in the property attributes of initial beam/shell models. As detailed designs evolve, equivalent properties in the global model need to be updated to account for cumulative and detailed modifications. Due to the different requirements of the various classes of simulation, procedures for abstracting simpler analysis models from detailed geometry are needed. Conversely, design by structural feature can help facilitate conceptual CAD. Idealised models typically implemented in preliminary design can be used to optimise equivalent 3D solids containing thin sheets, joints and long slender parts. This approach also helps in feature suppression, where artefacts below the scale of interest are removed from the analysis model. Simulation features are typically either regions of geometric complexity that are small relative to the environment in which they exist, or regions which have one or more dimension which is small compared to the others which can be represented with reduced-dimensional elements. An alternative representation of geometry known as the medial axis transform which facilitates the identification of structural features is described. For legacy models and those not designed with simulation in mind, tools for analysis feature recognition are required to enable appropriate simulation models to be identified for parts not designed with simulation in mind. Virtual topology, altering the topology of the geometric model without changing its underlying geometry, has been a key concept in facilitating the derivation of appropriate analysis models from design geometry. Issues relating to the use of Virtual Topology are addressed. One of the most difficult issues facing CAD-analysis integrations is how to update the properties of high-level global models to reflect the results obtained from detailed local models. Mixed dimensional modelling, used in combination with cellular partitioning of geometry and substructuring for linear problems, provides one possible way of managing the evolution of the design from a simple reduced dimensional model to a cellular solid. An updated global model can be recovered from dimensional reduction of the cells corresponding to thin sheets or slender bars and 3D models of connections substructured to equivalent joint stiffnesses using mixed dimensional coupling. The extension to nonlinear problems is obviously difficult but the general approach may have some merit. Whilst adaptive analysis to control discretisation error is well accepted, there has been very little work on the assessment the appropriateness and accuracy of a given simulation model. In this paper some simple assessments involving implied stress jumps between mixed dimensional models and between local and global models will be described. Open frameworks for collaborative, distributed design utilising heterogeneous, best-in-class modelling and simulation systems are identified as key technology for the future. Considerable effort has been expended on the development of analysis standards and communication transports. Utilities have appeared that allow seamless collaborations between multiple CAD systems over the web. XML, binary data formats and interfacing tools that are set to develop further and in parallel with emerging grid-computing technologies are also discussed.
| ||||||||||
