www.scorec.rpi.edu/Trellis

	Home SCOREC   About Trellis Documentation Code Documentation Projects/Papers   Downloads Support   SCOREC Publications AOMD
		Trellis Documentation The initial version of Trellis was developed by Mark Beall and documented in his Ph.D. thesis "An Object-Oriented Framework for the Reliable Automated Solution of Problems in Mathematical Physics". Even though Trellis has been developed further since then the thesis still does a good job explaining the fundamental ideas behind Trellis. The thesis and a short version in the form of a journal paper are available in pdf form. Trellis has some unique features such as direct links to geometry to support adaptive processes, and hierarchic variable order p-version shape functions. Trellis is easily extensible to other types of analyses than just Finite Elements. A different numerical method worth mentioning here is the Partition of Unity method, which has been implemented within Trellis. To support different types of numerical analysis capabilities the abstraction of the solution process was designed to be independent of the type of numerical technique used. Trellis represents the analysis process as a series of transformations of the problem from the original mathematical problem description to a set of algebraic equations. This transformation starts at the mathematical problem description level, which is transformed into a discrete system representing the discretized version of the model and attributes and the weak form of the Partial Differential Equation. The last transformation transforms the discrete system to an algebraic system, which corresponds to the process of calculating the individual contributions to the global system of equations. The discrete system views the problem in terms of contributions from a set of objects that are associated with the discrete representation of the model. These objects are called system contributors. The abstraction of the problem in terms of the system contributors is one of the more important abstractions in Trellis. It effectively decouples all the details of performing the low level computations (basis function, interpolation, geometry mapping, numerical integration, assembly and linear algebra) from the higher level computations of implementing a particular type of solution procedure (e.g. finite element) or a particular formulation. Even though Trellis is much more than just a Finite Element program many people will probably start using it by coding a particular Finite Element into Trellis. There are many elements already implemented and ready to go, from simple heat transfer and linear elasticity to higher order stabilized mixed elements for large strain/large displacement analysis of Hyper-Elastic Deformations and stabilized elements for flow problems. Those can serve as examples on how to create your own element. There is also this pdf document walking beginners through the steps. The idea of applying boundary conditions on the geometry instead of the Finite Element mesh is a major stone in the foundation of geometry based analysis. An attribute module to manage those boundary conditions as well as other type of information needed for a simulation has been developed together with a unit system. There are a few material models implemented in Trellis for which documentation is available. Follow this link to obtain more information.
				Ottmar Klaas