This thesis covers different methods of topology search for making time-domain simulation models based on geometric information. The application is in the field of fluid-power. Automatic methods for extracting simulation models from CAD data are presented.

A fluid-power component's performance is usually measured in quantities such as flow, pressure, and efficiency. The engineers designing such components or systems very often have to take special notice of space requirements, sonic noise emission and fatigue problems. Normally, these problems are analysed in a suitable program for the particular analysis. Space requirements are checked towards the geometry in the CAD-program. Noise-problems and fatigue are handled in FEM packages. Crafting a good simulation model that connects parameters in the fluid-power domain to the geometric characteristics of a CAD model requires a lot of timeconsuming work and a highly skilled engineer. This project, as presented in this thesis, has attempted to evaluate methods for making the model creation process automatic.

The research is presented as four papers that form a chronological report on the development of the search algorithms. The first two papers evaluate the three-dimensional application of well-known topology search techniques from the planar case. The third paper covers improvement to the topology search using native data from CAD programs. The last paper concerns the reduction of topologies, removing redundant information without losing the important connection between geometry and fluid-power phenomena.

Several examples are given in the application of topology search algorithms on fluid-power components, like valve-manifold and orifice identification in valves. Often the internal geometry of a valve-manifold includes a large number of free-form surfaces. This set of confining surfaces can be described as a complex cavity in most cases. Each face has several local properties involving neighbour distances, perimeter length and child entities. All of these properties can normally be found in the kernel functionality of modern solid modellers. To avoid large computer efforts in the topology search, the project have moved from using general approaches involving large number of intersection tests towards using kernel functionality and traversing the edges of the confining faces. The topology, as an abstraction tool, has come in naturally in the latter case. The thesis focuses on the understanding of topology search methods and not on the specific implementation issues.