To improve the efficacy and efficiency of the engineering product development process, we have developed a software framework that supports virtual prototyping: simulation of engineering system behavior at all desired levels of model fidelity, including the interactions of individual physical modalities. The framework also formalizes the specification and management of the computational processes for analyzing and improving the system designs, analysis processes to replace ad hoc, tool-specific scripts and interfaces. The framework achieves these capabilities by combining a modular dataflow architecture with an underlying integrated data model at a largely geometry-aware, and can be multi-physics, multi-disciplinary, multi-fidelity, and evolutionary.; We believe that the key to developing a software architecture that can fully support and exploit the use of virtual prototyping in the design of engineering systems is to bring together the following elements. (1) A which supports multi-physics analysis at all desired levels of fidelity. (3) A modular dataflow environment that manages the specification and execution of engineering analysis processes. (4) Interfaces to an extensible set of existing specialized tools, such as single-physics analysis and legacy black-box optimization codes. (5) Dynamic attachment of all data associated with a modular network. (a) Exposure of all data to dataflow operations. (b) Strongly typed, extensible module interfaces. (6) An interactive, extensible programming language, which gives the user the ability to define specify and drive engineering analysis processes. (7) Extensibility to new physics domains, analysis methods, and applications, through dynamic typing, the composition of existing ones.; Our contention is supported through the extension of an existing software framework, CoMeT, to incorporate all of the elements listed above. The efficacy of the system is demonstrated by multi-physics, multi-fidelity analysis and control of a simulated membrane telescope.
展开▼
