Automated Modeling Research Topics
The Automated Modeling research program is focused on the technologies needed to support the effective application of simulation by engineers and scientists. Current SCOREC activities in this area include:
- Simulation Environment for Engineering Design
- Automatic generation, modification and control of meshes
- Interoperable mesh and discretization technologies
- Model Generation and Control for Multiscale Modeling
Simulation Environment for Engineering Design
In engineering design, simulation technologies must provide reliable estimates of performance parameters and sensitivity information needed to support decision processes. Activities that should be supported range from the manufacturing process optimization, to a surgeon interactively using patient specific simulations to develop a treatment plan.
For simulation to be used as a design tool it must be accessible and usable by designers. Currently, CAE specialists who understand the intricacies of the current tools are needed to perform the desired simulations. However there is a limited supply of these CAE experts and it is not economical for a corporation to hire and train a sufficient number to carry out simulation on the scale that is needed. A recent D.H. Brown report estimated that it takes 5 years and $400,000 to train an effective CAE expert.
The only viable solution is for simulation technologies that work within the designer's environment, providing them the information they need. However, this alone is not a sufficient solution. These designers do not have the level of understanding of numerical analysis that a CAE expert has and are not in a position to evaluate whether the results of a simulation are sufficiently reliable to base their decision processes. Thus, the system must be able to evaluate and automatically control the simulation process to ensure that the designer is provided results at the level of accuracy requested.
The input provided for a simulation includes the design as it currently exists, the set of performance parameter values and/or sensitivities to be estimated, and level of accuracy required. To calculate the requested information the simulation environment must:
- Support the interactions of the product definition with the simulation technologies.
- Determine the appropriate sets of models for the simulations.
- Automatically generate the numerical discretizations that constitute the simulations.
- Assess the accuracy of the predicted parameter values and/or sensitivities and improve the simulation models as necessary to obtain the accuracy desired.
To date, efforts to bring simulation into the design/manufacturing process have focused on individual tools that operate only on descriptions natural to the CAE tools used. The lack of technologies to automatically execute reliable simulations from the product definition is a major reason simulation is not used early in the design process. SCOREC is working with Simmetrix Inc. on a NIST ATP project to provide a set of functions and underlying structures that are substantially richer than those used by current CAE technologies and, through the appropriate integration with PDM, CAD, and CAE technologies, create a Simulation Environment for Engineering Design (SEED). As part of this project the Simmetrix/SCOREC team are working with several industrial partners on the development of simulation-based design systems for specific application areas.
The figure below indicates the overall structure of SEED. Within the dotted box are the five key components needed for the effective integration of simulation into the industrial design/manufacturing processes. The basic functions of these five components are:
- Simulation Model Management: Responsible for interactions between the product design as defined in the product management system and the simulation technologies. The structures and methods must support simulation processes to the lowest levels, while reflecting simulation dictated design changes to the highest level of the product representation.
- Simulation Data Management: The simulation procedures use various discretizations of the product domain and produce information defined at that level. Structures and methods are needed to define this information and properly relate it to the product representation.
- Simulation Model Generation Tools: Performs the geometric operations to construct models and the appropriate domain discretization for a simulation from the product definition.
- Adaptive Control Tools: Responsible for determining the appropriate mathematical models, selecting discretization technologies, evaluating the accuracy of the predictions, and determining the improvements needed to obtain the desired accuracy.
- Geometry-Based Simulation Engine: Responsible for executing the numerical aspects of the simulation procedures. It must effectively support adaptive procedures and support the ability to link with existing CAE tools.
Automatic Generation, Modification and Control of Meshes
SCOREC has been involved with the development of automatic mesh generation and modification technologies for a number of years. The tools that have been developed support the automatic creation and modification of meshes directly from non-manifold solid models. The tools make use of the flexible object-oriented structures representing the geometric model, attribute information associated with the model, and the mesh itself.
boundary layer mesh crack propagation
coarse h-mesh & curved p-mesh
The developed procedures have been used by a number of sponsors over the years and have formed the basis of commercial automatic mesh generation products. Most current SOCREC efforts in these areas are focused on particular capabilities that remain to be fully developed including:
- Ensuring mesh refinement procedures are able to place node points on the curved boundary of the model domain. For more information see the linked PowerPoint presentation.
- Generation and control of proper curved meshes for use when high order discretization methods (e.g., p-version finite elements) are used to analyze a problem on curved domains. For more information see the linked PowerPoint presentation.
- Perform intelligent mesh modification to account for the needs to mesh size changes (refinement/coarsening) and control of mesh entity shape in evolving geometry problems. Efforts are underway to have these procedures operate to satisfy the general anisotropic mesh metric fields. Since these procedures are often applied to simulations where solution fields must be transferred between meshes during mesh modification, a set of incremental solution transfer procedures are also being developed. For more information on the basic mesh modification procedures see the linked PowerPoint presentation.
Interoperable Mesh and Discretization Technologies
SCOREC is a major partner in the Center for Terascale Simulation Tools and Technologies (TSTT) that is a part of the Department of Energy's Scientific Discovery through Advanced Computing (SciDAC) program. The TSTT partners include Argonne, Brookhaven, Lawrence Livermore, Oak Ridge, Sandia, and Pacific Northwest National Laboratories, Rensselaer and the State University of New York at Stony Brook.
A summary of the TSTT project goals, as provided by the proposal abstract, is: "The primary objective of the Terascale Simulation Tools and Technologies (TSTT) center is to develop technologies that enable application scientists to easily use multiple mesh and discretization strategies within a single simulation on terascale computers. We will focus our efforts in the areas of high-quality, hybrid mesh generation for representing complex and possibly evolving domains, high-order discretization techniques for improved numerical solutions, and adaptive strategies for automatically optimizing the mesh to follow moving fronts or to capture important solution features. We will encapsulate our research into software components with well-defined interfaces that enable different mesh types, discretization strategies, and adaptive techniques to interoperate in a "plug and play" fashion (see figure below). All software will be designed for terascale computing environments with particular emphasis on scalable algorithms for hybrid, adaptive computations and single processor performance optimization. To ensure the relevance of our research and software developments to the SciDAC goals, we will collaborate closely with both SciDAC application researchers and other ISIC centers. In particular, we will insert existing TSTT technologies into fusion, accelerator design, climate modeling, and chemically reacting flow applications, and thereby have both significant near-term impact on those efforts as well as receive the feedback necessary to ensure our success. We will show the potential long-term impact of TSTT center research by deploying hybrid solution strategies in SciDAC applications and demonstrate the ability of researchers to easily obtain superior simulation results."
SCOREC's focus in TSTT is the development of the methods needed to support terascale computing on adaptively refined meshes and methodologies to support the interoperability of various numerical analysis and mesh generation technologies. The SCOREC developments on this project will be freely available to the community as open source code. This will include the serial and parallel algorithm-oriented mesh database and a geometry-based discretization library focused on adaptive high order methods (Trellis). SCOREC is also heavily involved in the support of the development of simulation tools for SciDAC applications including fusion and astrophysics.
Model Generation and Control for Multiscale Modeling
Multiscale simulation procedures employ various sets of analysis discretizations that must be related in the scale linking processes. To support these operations a discrete model tool is being developed that constructs a complete non-manifold topological model for domains that are already discretized by some method. The procedures under development build on SCOREC mesh generation technologies and tools that support discrete level simulations (voxel and octree decompositions). They are being used to link atomic level to continuum, and continuum micromechanical to continuum macromechanical models. For more information on linking atomic and continuum modeling see the linked PowerPoint presentation.