Report

[email protected] Center for Computation and Technology Visualization of finite element data of a multi-phase concrete model M. Ritter1, M. Aschaber1, W. Benger2, G. Hofstetter1 1 University of Innsbruck, Austria 2 Louisiana State University, USA 10.7.2013, Vienna • Outline – Motivation – Numerical Simulation – Data Modeling – Visualization – Conclusion & Future Work Motivation • Motivation: Scientific Visualization Techniques and Research Engineering Simulation Tools and Visualization by Neal Stone Gap Motivation • Motivation: Scientific Visualization Techniques and Research Engineering Simulation Tools and Visualization by Neal Stone Motivation • Motivation: • Simulation techniques are used more frequently • Produced data sets growing • Data complexity is increasing • Visualization used for data interpretation of results is important. 10.7.2013, Vienna Numerical Simulation • Aim: – More realistic simulation of drying shrinkage • Application: – Strengthening of a RC structure by adding an overlay New top concrete layer Old concrete structure Numerical Simulation • Drying shrinkage: – – – – Long term drying process in concrete Decrease of relative pore humidity Increase of capillary pressure Capillary pressure results in volumetric shrinkage drying concrete Numerical Simulation • Drying Shrinkage: New top concrete layer Old concrete structure Different internal stresses Critical region at joint Drying shrinkage Swelling concrete Numerical Simulation • Numerical Simulation: – Finite element simulation on multiple grids of concrete specimen – Hexahedral Mesh of 9 x 9 x 13 Cells 100 x 100 x 56 mm Numerical Simulation • Numerical Simulation: – Finite element simulation on multiple grids of concrete specimen – Hexahedral Mesh of 9 x 9 x 13 Cells Undeformed linear element Deformed quadratic element The element has curved faces Numerical Simulation Numerical Simulation: • Multiphase concrete model • Solid, water, and gas phase (dry air and water vapor) • Coupled hygral-thermo-mechanical model Governing equations: • Balance equations • Mass • Enthalpy • Linear Momentum • Linear kinematic relations • Constitutive equations Multiple solution variables (Data fields) • Gas pressure • Capillary pressure • Displacements • Temperature scalar scalar vector scalar Derived data fields • Strain • Stress 2nd order tensor 2nd order tensor Numerical Simulation • Drying shrinkage εℎ – Effective stress: – Hydrostatic pressure of the water on the solid phase: 10.7.2013, Vienna Data Modeling Before doing data visualization one has to deal with data • Many different kinds • Many formats Data management and handling is crucial in computational sciences • Reusability of methods and techniques • Sustainability • Exchangeability of data (collaborations) We propose using a concept based on mathematics to systematically organize data Data Modeling Fiber Bundle Data Model Inspired by concepts of: Topology Differential Geometry Geometric Algebra Separation of Geometry (Grids) and Datafield (Fields) Grid Field Data Modeling Grid the base space • Manifold describing the base space • Topology • Refinement level • Coordinate representation • Vertex positions in representation • Neighborhood Data Modeling Field the fiber space • Dataset holding numerical data • per k-cell on the grid (vertex, edge, cell, … ) • Array of arbitrary type, for example: • Scalar • Vector, BiVector, … • Tensor • Any other user defined type Data Modeling • Hierarchical structure: Data Modeling Supported Grid types: • • • • • • • Uniform Grid Curvilinear Grid Rectilinear Grid Adaptive Mesh Refinement Grid (AMR) Point Cloud Lines Triangular/Quad and Mixed Surfaces Grids can be fragmented (Blocks) having Ghost Zones Grids can have refinement levels Work in progress: • • • • Hexahedral Grid FEM Grid Connected Graph Data Full Waveform LIDAR Laser Data Data Modeling Data Modeling • Data at Vertices: (optional) (optional) Data Modeling • Data at Integration Points: No positions can be computed Data Modeling • Sets of Nodes and Sets of Elements: Indices of vertices in named fragments Indices of integration points in named fragments Data Modeling • Linked groups for alternative data access: – E.g. time frames and time steps in ABAQUS Li nk Bundle - frames Bundle - steps T=0.0 T=0.0 Link T=0.2 T=1.0 T=0.4 T=0.6 T=0.8 T=1.0 T=1.2 T=1.4 No data stored Data Modeling FEM - Example HDF5 Based Data format: independent, free, open, data browser www.hdfgroup.org 10.7.2013, Vienna Visualization Visualization Shell VISH • Visualization framework • Highly modular design – Small Core – Plug-Ins • Mainly developed by Werner Benger – Currently about 8 people are actively contributing • • • • C++, OpenGL Open Academic License Runs on Linux, Windows (and MacOS) http://vish.fiberbundle.net Visualization Visualization Visualization • Colored Cages – Show FE grid • Positive • Negative – – – – – Shaded colored surface Illustrates data at vertices One scalar field via color-map One vector field via displacement Can be combined with other visualization techniques Visualization • Colored Cages – Integration point data is extrapolated and averaged on demand Extrapolation from integration points Over-scaled deformation Averaged, smoothed Visualization • Tensor analysis: – Shape factors by [Westin97] – Stress/strain are 3x3 symmetric tensors – 3 Eigen-Values: – Shape factors: [BBHKS06] Visualization • Direct stress tensor visualization: – Ellipsoids representing the shape factors – Tensor Splats [BengerHege04] -> barycentric Visualization • Direct stress tensor visualization: – Ellipsoids representing the shape factors – Tensor Splats [BengerHege04] Works only for positive Eigenvalues! -> barycentric Visualization • Direct stress tensor visualization: – Ellipsoids representing the shape factors – Tensor Splats [BengerHege04] Works only for positive Eigenvalues! Enhancement: Color a splat in blue, when any Eigenvalues is negative -> barycentric Visualization • Drying simulation: – Tensor splats: Biaxial tension Uniaxial tension Pressure region + Multiple stress directions + Tension vs pressure regions Visualization • Scalar fields by volume rendering: – Texture based volume rendering ( requires resampling on uniform grid to be improved) – Shows inner structure of data fields – Example: tri-axial compression of a cuboid σ and Mises in ABAQUS σ and Mises via volume rendering • Drying simulation: – After 30 days of drying – Volume rendering of drying shrinkage εℎ – Cages show an uplift of the corner and the edges • Dual volume rendering: – One scalar field controls color – Another controls transparency 10.7.2013, Vienna Fiber bundle data model for FEM • Captures many other types of scientific data • Comes with an HDF5 based data format (big data) Good for collaborations and transparent data storage Visualization: • Colored cages • Direct tensor field visualization • Scalar fields via volume rendering • Dual volume rendering Future work: • Enhancing the direct tensor field visualization • Volume rendering on the FEM grid (GPU raycasting) • Support more FEM data (also shells) 10.7.2013, Vienna Numerical Simulation • Balance equations of the multiphase model Mass of the water w Mass of the steam gw Mass of the dry air ga Mass of the solid phase s Enthalpy of the whole system Impulse of the whole system