SIMULIA / Abaqus / Engineous - ein Update
Martin Küssner Dassault Systemes Simulia GmbH
[email protected]
Content • Integration of processes • Multiphysics • Unified FEA • Parallelization • Material laws • Simulation Lifecycle Management • Optimization/Robust Design • Conclusion
Integration of processes „FEA is not all that is out there ….“
Automotive
Innovation Integration Process Centered on Virtual Experience
Aerospace
Shipbuilding
Industrial Equipment
Consumer Goods
Consumer Packaged Goods
Life Sciences
High Tech
Construction
Energy
Business Services
Multiphysics „There is more that matters than just stresses and strains“
Multiphysics in Abaqus Unified FEA •
Abaqus enables coupling of multiple fields Thermal
Electrical
Acoustics
Structural Structural
Piezoelectrical
Courtesy: Honeywell FM&T
Fluid flow
Courtesy of Dr. Michelle Hoo Fatt (University of Akron)
Pore pressure
Coupled Eulerian-Lagrangian (CEL) • Enhances the ability of Abaqus/Explicit to model fluids and materials that undergo extreme deformation Capability consists of: A three-dimensional Eulerian element type Support for all materials in Abaqus/Explicit except hyperelasticity and anisotropy General contact between Eulerian and Lagrangian domains Parallel processing Target applications include: Hydroplaning, water impact, earth penetration, sloshing, bird strike Fully supported in Abaqus/CAE
Volume fraction tool being used to fill a bottle for a drop test
Coupled Eulerian-Lagrangian (CEL) • Uses a multi-material finite element formulation • Volume-of-fluids method tracks material boundaries within an Eulerian domain • Conforming meshes are not required Provides a specialized technique to simulate FSI problems that include: Complex structural contact conditions including self-contact Large structural deformations and displacements Very high-speed dynamics Damage, failure, or erosion at the fluid-structure interface
Coupled Eulerian-Lagrangian (CEL) • Example: Hydroplaning • When a vehicle is driving in the rain, water trapped between the road and tire leads to pressure build-up and possible loss of traction • Tire manufacturers use simulation to design tread patterns that reduce the possibility of hydroplaning • CEL easily handles the complex contact conditions (“pinching”) that occur during hydroplaning simulation
Unified FEA „Re-Using instead of Re-Doing“
Implicit/explicit integration • Unified FEA™ is a key component of our product strategy • Implies the ability to easily transfer models and results between implicit and explicit solution technologies • Implicit/explicit integration • Most elements and materials that are common to both solution technologies can be transferred
11
Nonlinear Dynamics • Abaqus/Standard (=Implicit) • Uses a second-order accurate, implicit scheme called the HilberHughes-Taylor (HHT) rule. • This is a generalization of the Newmark method. • Second-order accurate means the scheme integrates a constant acceleration exactly. • The method is unconditionally stable: any size time increment can be used, and the solution will remain bounded. • Abaqus/Explicit (=Explicit) • Uses a second-order accurate, explicit integration scheme. • The method is conditionally stable—it gives a bounded solution only when the time increment is less than a critical value.
Comparing implicit and explicit integration scheme Implicit • Time increment size is not limited: generally fewer time increments required to complete a given simulation. • Each time increment is expensive since each requires the solution for a set of simultaneous equations.
Explicit • Time increment size is limited: generally many more time increments are required to complete a given simulation. • Each time increment is relatively inexpensive because not required to solve a set of simultaneous equations. – Most of the computational expense is associated with element calculations (forming and assembling I).
Seek to control the residual at an intermediate point The model is in equilibrium at the beginning and end of the increment
Implicit • Ideal for problems where the response period of interest is long compared to the vibration frequency of the model. • Difficult to use explicit dynamics effectively because of the limit on the time increment size. • Use for problems that are mildly nonlinear and where the nonlinearities are smooth (e.g., plasticity). • With a smooth nonlinear response Abaqus/Standard will need very few iterations to find a converged solution.
Residual at half-increment (half-step residual)
Explicit • Ideal for high-speed dynamic simulations – Require very small time increments; implicit dynamics inefficient. • Usually more reliable for problems involving discontinuous nonlinearities. – Contact behavior is discontinuous and involves impacts, both of which cause problems for implicit time integration. – Other sources of discontinuous behavior include buckling and material failure.
Abaqus/Standard • Comprehensive linear & nonlinear implicit general purpose finite element analysis of structural, thermal, acoustic & mechanism simulations • Integration with Abaqus/Explicit provides maximum flexibility for multi-physics simulation (“Unified FEA”) • Sophisticated contact, failure, material modeling & other advanced nonlinear capabilities • High-performance direct and iterative solvers with support for shared and distributed memory configurations • Powerful interfaces for user customization
Abaqus/Explicit • Comprehensive explicit finite element analysis of structural, thermal, acoustic & mechanism simulations • Integration with Abaqus/Standard provides maximum flexibility for multi-physics simulation (“Unified FEA”) • High-performance solver with support for shared and distributed memory configurations • Powerful interfaces for user customization
Courtesy BMW
“Abaqus Unified FEA” • Abaqus import capability can transfer a model or part of a model, together with associated state information, between an Abaqus/Explicit analysis and an Abaqus/Standard analysis
Abaqus/Explicit
Abaqus/Standard
Abaqus/Standard
Abaqus/Explicit
Abaqus/Explicit
Abaqus/Standard
Overlapping material library • Damage and failure of fiber-reinforced composites • Example: Hashin’s damage initiation criterion of unidirectional composites is available in Abaqus/Standard and Abaqus/Explicit • The model captures four different damage mechanisms • Fiber rupture • Fiber buckling and kinking • Matrix cracking • Matrix crushing • Damage evolution consistent with damage framework introduced in Version 6.5 • Abaqus/Explicit ↔ Abaqus/Explicit and Abaqus/Standard ↔ Abaqus/Explicit import
Example (Payload Fairing) • Pressure load = 20 x (cos2θ) psi under side only
Note : strain discontinuities in doubler region
Example (Payload Fairing) • Composites supported in both /Standard and /Explicit solver /Standard (20 psi)
/Explicit (20 psi)
Example (Payload Fairing) • Composites supported in both /Standard and /Explicit solver /Standard (20 psi)
/Explicit (20 psi)
Example (Payload Fairing) • Analysis in /Explicit carried out to collapse of structure Applied Load = 50 psi over 1.0 second
P = 42.5 psi (t = 0.85)
Nose Tip Displacement
Parallelization „We need to become MUCH faster ….“
Parallelization and performance • DMP direct sparse solver • Scalability significantly improved • Memory use noticeably reduced when running on 4 or more compute nodes • Unsymmetric solver now supported (important for contact simulations) • Multiple linear loads cases and Riks method now supported • DMP is an effective strategy for Abaqus/Standard Courtesy of the DANA Corporation
Major Breakthrough in Abaqus Performance • Expanded procedure support • Improved scalability • Reduced memory requirements Courtesy DANA 9.2M DOF
92% Speed Up
Material Laws „Things became really difficult ….“
Material law
„The calibration of the second invariant of the deviatoric left CauchyGreen tensor causes troubles when calibrating hyperelastic material law.“ What to do with these kind of information?
Simulation Lifecycle Management „What to do with all the data and how to store the processes….?
Simulation Lifecycle Management • SLM means: • Bringing Order to Simulation: • Management of Data, Methods and Processes • Connecting users to each other and the enterprise • Leveraging Simulation IP: • Capturing simulation know-how and related decisions • Extracting and re-using the built-in value of simulation activity • An Open Platform to Manage and Deploy Applications: • Workflow chaining and job submission • Connector architecture for 3rd party applications
SLM Sneak Preview
5 3
2
4 1
1. 2. 3. 4. 5.
3DSearch 3DCompass 3DNavigation 3DHeads-Up Contextual Buddy List
Optimization / Robust Design
Isight for Abaqus: Benefits • Reduce design time • Execute multiple simulation studies automatically overnight • Parallel submission of optimization, Monte Carlo, and DOE jobs on multiprocessor machines or in conjunction with LSF • Improve quality • Design-to-target for simulation attributes • Account for variation in materials, loads, tolerances, and operating conditions • Find the best design • Understand which model parameters drive targets • Trade off design alternatives in real time with colleagues and other stakeholders
The Big Picture – Automated Design Strategies “Shop” for the best design Y2
Outputs
Constraint Boundary
DOE: Critical Factors and Initial Design X2
Initial Best Design
Inputs
Feasible (safe)
Infeasible (failed) Y1
X1
Isight for Abaqus: Design of Experiments (DOE) Perform trade-offs and understand the design space •
•
Capabilities •
Determine which input variables have the most influence on your simulation outputs
•
Use to build approximation models
•
Estimate of an Optimal Design
Types •
Parameter Studies
•
Orthogonal Arrays
•
Full Factorial Arrays
•
Optimal Latin Hyper Cube
•
Latin Hyper Cube
•
Central Composite
•
Import Outside Experiments
Isight for Abaqus: Approximations Build from DOE samples, speed up calculations from Six Sigma, speed up live performance tradeoffs • Types • Response Surface Model (to 4th order) • Radial Basis Function (RBF)
• Capabilities • Automatic setup • Automatic error estimation • Effects graphs • Interactive tradeoffs • Simulation Surrogate
The Big Picture – Automated Design Strategies
Y2 Improve Design Performance
Outputs Optimized Design Constraint Boundary
Optimization (Approximations)
Initial Design from DOE Feasible (safe)
Infeasible (failed) Y1
Isight for Abaqus: Optimization Drive toward a target performance • Capabilities • Formulate variables, constraints, and multiple objectives • Multi-objective Pareto fronts
• Types
Sim approximation surrogate
• Gradient: NLPQL • Multi-Objective: NCGA, AMGA • Pattern: Hooke-Jeeves and Downhill Simplex • Exploratory: Multi-Island Genetic Algorithm (MIGA), Adaptive Simulated Annealing (ASA) • Automatic Optimization: Pointer Automatically configures NLPQL, an evolutionary algorithm, Downhill Simplex, and a linear solver
The Big Picture – Automated Design Strategies
Improve Design Quality
Y2
Outputs % Reliable
% Unreliable
Constraint Boundary Design for Six Sigma Robustness and Reliability Analysis and Optimization
Robust and Reliable Design
Feasible (safe)
Infeasible (failed) Y1
Conclusion
SIMULIA / Abaqus / Engineous - ein Update
Martin Küssner Dassault Systemes Simulia GmbH
[email protected]
