1 / 32

Nonsmooth DEM Simulation Fast and stable simulation of granular matter and machines

Nonsmooth DEM Simulation Fast and stable simulation of granular matter and machines. Dr C. Lacoursière 1 , Dr Martin Servin 1 , A. Backman 2 1 Umeå University, 2 Algoryx Simulation 2010 - 08-24. Fast and stable simulation of granular matter and machines. Targeting

reyna
Download Presentation

Nonsmooth DEM Simulation Fast and stable simulation of granular matter and machines

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Nonsmooth DEM Simulation Fast and stable simulation of granular matter and machines Dr C. Lacoursière1, Dr Martin Servin1, A. Backman2 1 Umeå University, 2 Algoryx Simulation 2010-08-24

  2. Fast and stable simulation of granular matter and machines • Targeting • Simulation of complex mechanical systems • - short computing time e.g. realtime • Main results • Unified method – constraint based multibodies • Direct-iterative hybrid solver fornonsmooth DEM • Simplification & parallelization •  fast and stable simulation

  3. Fast and stable simulation of granular matter and machines • Realtime simulators for training, marketing and R&D, operator-in-the-loop • 100-1000 constrained multibodies running in 60 Hz • time-budget per update < 0.02 s

  4. Two alternative strategies • Smooth DEM (Cundall) • Δt < τelast • explicit force models (Hertz) • small time-step, explicit integration • Nonsmooth DEM (Moreau/Jean/Renouf/Anitescu) • Δt > τelast • implicit contact laws (unilateral constraints & impulses) • large time-step, implicit integration Radjai

  5. Summary of the method • Constraint based modeling • Constraint regularization • Variational time-integration of DAE system • Linearization into MLCP • Block-sparse direct-iterative hybrid solver

  6. Summary of the method • Constraint based modeling • system of particles and rigid bodies • constraint types: nonpenetration, dry friction, joints, motors, incompressibility • Lagrange multiplier - auxiliary variable – inequality, complementarity conditions • Constraint regularization • Variational time-integration of DAE system • Linearization into MLCP • Block-sparse direct-iterative hybrid solver

  7. Summary of the method • Constraint based modeling • Constraint regularization • stabilization • model elasticity and viscosity • Variational time-integration of DAE system • Linearization into MLCP • Block-sparse direct-iterative hybrid solver

  8. Summary of the method • Constraint based modeling • Constraint regularization • Variational time-integration of DAE system (Marsden) • SPOOK integrator (Lacoursière) • similar to Rattle and Shake … plus inequality and complementarity conditions • Linearization into MLCP • Block-sparse direct-iterative hybrid solver

  9. Summary of the method • Linearization into MLCP • Mixed linear complementarity problem • Fluid: 150K x 150K • Rocks:2K x 2K • Loader: 150 x 150

  10. Summary of the method • Constraint based modeling • Constraint regularization • Variational time-integration of DAE system • Linearization into MLCP • Block-sparse direct-iterative hybrid solver • SuperLU • linear scaling • Fluid: 150K x 150K • Rocks:2K x 2K • Loader: 150 x 150

  11. Accelerating the computations • Splitting the system – direct-iterative hybrid solver • Parallelization • Merge/split simplification • Contact reduction

  12. Accelerating the computations • Splitting the system – direct-iterative hybrid solver • joints and non-penetration (direct solver) • joint limits and friction (iterative solver) • Parallelization • Merge/split simplification • Contact reduction

  13. Accelerating the computations • Splitting the system – direct-iterative hybrid solver • Parallelization • constraint based SPH fluid on GPGPU using precondition conjugate gradient • inequality constraints (dry friction) difficult • Merge/split simplification • Contact reduction

  14. Accelerating the computations • Splitting the system – direct-iterative hybrid solver • Parallelization • Merge/split simplification • Contact reduction

  15. Accelerating the computations • Splitting the system – direct-iterative hybrid solver • Parallelization • Merge/split simplification • Contact reduction • eliminate redundant contacts

  16. Results and examples • Wheel loader system • Constraint fluid (SPH) on GPGPU • Rotating drum with rigid and fluid elements • Pellet balling plant

  17. Results and examples • 12 rigid bodies, 15 joints • 500 rigid rocks • 5000 contacts • 0.02 s time-step • 1 s simulation = • 0.1—3.5 s computation • 75% solver • 20% collision detection • 5% merge/split • 2–10 speed-up • Laptop: 64 bit, 3.03 GHz. 4GB RAM • Wheel loader system

  18. Results and examples • Wheel loader system • Shmulevich (2007) • Validation system – measuring blade force and particle displacement

  19. Results and examples • Wheel loader system • Constraint fluid (SPH) on GPGPU • Rotating drum with rigid and fluid elements • Pellet balling plant • SPH incompressibility • constraint • 60 l constraint fluid • 0.01 s time-step • 60 l standard SPH • 0.01 s time-step

  20. Results and examples • 0.9 M particles • 0.001s time-step • density of water • 4 wide • conjugate gradient • 1 s simulation = • 10 s computation • Speed-up • 10x - incomprconstr • 100x - GPGPU • Constraint fluid (SPH) on GPGPU • NVIDIA GTX 280 graphics card • Hosted by Intel Core i7 Processor

  21. Results and examples • Wheel loader system • Constraint fluid (SPH) on GPGPU • Rotating drum with rigid and fluid elements • Pellet balling plant Extension of cosstraint SPH toviscoplastic fluids for complex materials, e.g., size mixtures, wet material • Bui, Int. J. Numer. Anal. Meth. Geomech. 2008; 32:1537–1570 • Jop, Nature, Vol 441|8 June 2006

  22. Results and examples • Wheel loader system • Constraint fluid (SPH) on GPGPU • Rotating drum with rigid and fluid elements • Pellet balling plant • rigid drum driven with • motor constraint • 222 rigid bodies • 1235 SPH fluid particles • 860 contacts • 0.02 s time-step • 1 s simulation = • 10 s computation • non-penetration constraint • produce buoyancy

  23. Results and examples • Wheel loader system • Rotating drum with rigid and fluid elements • Constraint fluid (SPH) on GPGPU • Pellet balling plant

  24. Balling plant In: - fines (ore + binding) - undersized pellets 100 ton/h • Mono-sizedspherical pellets • Understand flow patterns and forces • Design parameters • Control parameters • Agglomeration process • 100K – 100M particles • Time scales 1ms – 5min • 1/10 fraction real vs simulated time Out: - green pellets Return: - undersized pellets - oversized pellets (crush)

  25. Balling plant Next step • Study of outlet design … – 2011-03 • PhD student project 2010 – 2014

  26. Validation and analysis

  27. Validation and analysis

  28. Validation and analysis

  29. Summary and conclusions • A unified method for combining nonsmooth DEM, SPH fluid and machines based on rigid multibodies • Constraint based approach  large time-steps, fast and stable  suitable for real-time simulators & efficient off-line simulation • A number of opportunities to further accelerate the computations • It is not well established what is lost in the nonsmooth approximation (Δt > τelast) – analysis is on-going • Suggestions for critical validation tests are much appreciated!

  30. Project proposal - demonstrator Demonstrator for simulation based design optimization and control of systems with granular material – nonsmooth DEM Suitable system for demonstrator: • Crushing, loading, sorting, conveying • Rigid body model for large elements (0.01 – 1 m) • Fluid model for finer material (100 um – 10mm) • Rock, mineral ore, biomaterials, dry/wet Gathering interested participants: • Plant owners – LKAB, Boliden, energy companies • System manufacturer –conveors, chrusher, control systems • Consulting – control and optimization • Simulation software companies – Algoryx, DEM Solution?

  31. Videos Realtimeconeyor with loader and rocks http://www.algoryx.se/~servin/vids/conveyor_06.avi Realtime simple balling drum http://www.algoryx.se/~servin/vids/rulltrumma_2.avi Robot conveyor demo http://www.algoryx.se/~servin/vids/AGX_robotics1.avi Constraint fluid http://www.algoryx.se/~servin/vids/constraintFluidsXVID.avi GPGPU fluid 1M particles at 10 Hz http://www.algoryx.se/~servin/vids/gpufluid1.avi GPGPU fluid + boxes http://www.algoryx.se/~servin/vids/constraint%20fluid%20with%20boxes.avi Viskoelastic fluid http://www.algoryx.se/~servin/vids/visko-elasto-plastic.avi Merge+split for realtime loader demo http://www.algoryx.se/~servin/vids/loader.wmv

  32. Links Research group Modeling and simulation of complex mechanical systems http://www.physics.umu.se/english/research/statistical-physics-and-networks/complex-mechanical-systems/ Algoryx Simulation - Spin-off company for prof simulation software development, 14 employees https://www.algoryx.se/, training simulaors: http://www.oryx.se/ UMIT Research Lab at Umeå University – Multidisciplinary research environment with 25+ researchers on computational science and engineering http://www.org.umu.se/umit/umit/?languageId=1 ProcessITInnpovaions – Innovation system connecting processing industry in northern Sweden with research and ITC and software companies http://www.processitinnovations.se/default.aspx?id=1765

More Related