Fracture Modeling in Computer Graphics A survey MIIACS

Fracture Modeling in Computer Graphics A survey MIIACS Lien Muguercia Torres Advisors Dr. Gustavo A. Patow Dr. Carles Bosch 1

Introduction 2

Introduction 3

Introduction 4

Why fracture? Extended application area No ideal solution This thesis 5 No specific survey

Overview Background Physically-based methods Non-physically based methods Conclusions Future work 6

Background 1 2 3 4 5 7 • Aging and weathering processes in CG • Cracks and fractures in nature • Physical models • Time integration

Aging and weathering processes External factors Internal factors Chemical attacks Mechanical damages 8 Biological processes

Aging and weathering processes 9 [Merillou 08]

Aging and weathering processes Specific aging model Simulation Global aging models Aging as particles [Chen 05] Example based texture synthesis [Gu 96; Matusik 03] 10

Cracks and fractures in nature Cracks: long and tiny Fractures: desattachments Bark Stone Brittle materials Dry cly Ductile Glassmaterials Wood Mechanic classifications Ice 11 Eggs

Cracks and fractures in nature Materials parameters Stress Strain Yield strength Fracture Trans-granular Inter-granular 12

Simulation fracture Common steps over time Compute internal forces Determinate location and orientation Modify the model Compute stress-strain 13 Determinate location Re-meshing

Physical models Mass-spring system Fast implementation/runtime Poor visual results FEM Simulation precision Computational cost Mesh-less methods Avoid mesh reconstruction Bounding conditions 14

Time integration Explicit Compute the state by extrapolating the previous one Methods Semiimplicit Solving the implicit methods by linearization 15 Implicit Compute the state at the end of the time step

Overview Background Physically-based methods Non-physically based methods Conclusions Future work 16

Physically-based methods 17 1 • Mass-spring 2 • FEM 3 • Mesh-less 4 • Other

Mass-spring models Rigid bodies Bodies as hybrid between rigid and deformable [Terzopoulos 88; Norton 91] Model behavior of solid objects 18

Mass-spring models Rigid bodies Bodies as hybrid between rigid and deformable [Terzopoulos 88; Norton 91] Model behavior of solid objects Voxels Connected with strong link values [Mazarak 99] Scaled/displaced in any directions [Martins 01] Has unique material properties 19

Mass-spring models Tetrahedral elements Each element equivalent six springs [Smith 00] Elements with same size cause undesirable artifacts [Hirota 00] and lead with size adapted [Aoki 04] 20

Finite element methods Classic Brittle/Ductile fracture animation [O’Brien 99; O’Brien 00; O’Brien 02] Simulation determinates cracks initiation and propagation analyzing stress value High computation time 21

Finite element methods Classic Fluid dynamics model based using O’Brien approach [Yngue 00] An approach using heuristical stress [Iben 09] Replace as few tetrahedral as possible [Wicke 10] 22

Finite element methods Hybrid Alternate rigid body and continuous model at the point of impact [Muller 01; Molino 04] 23

Finite element methods Hybrid Alternate rigid body and continuous model at the point of impact [Muller 01; Molino 04] Bi-layered materials Cracking induced by the material growth/shrinkage [Federl 02] Delaunay triangulation for mesh construction 24

Meshless methods Based on particles/point-based representation Compute spatial derivatives of displacement Synthetize crack surfaces as triangle meshes [Muller 04; Pauly 05; Steinemann 06] 25

Other methods Crack propagation based on multi-layer Cellular Automata [Gobron 01] Crack propagation by systematic stress release of unstable cells 26

Physically-based methods Which to choose? Mass-spring, simple and fast / quality FEM, accurate simulation / computational cost Mesh-less, avoid mesh treatment / bounding restrictions General problems Computational time vs. quality 27

Overview Background Physically-based methods Non-physically based methods Conclusions Future work 28

Non-physically based methods 1 2 29 • Image-based • Procedural

Image-based methods Information extracted from images → Textured height field pattern [Wang 03] Reproducing input lines from images [Mould 05] Mapping and Bump mapping with real images [Hsien 06] 30

Procedural methods Open, flexible and parameterizable solution Parallel strips to simulate bark generation [Lefebvre 02] Tools to control the patter by observation [Martinet 04] Mathematic algorithm to cracks in wax painting [Wyvill 04] 31

Procedural methods Connected voxel method as basis [Taubman 04; Valette 07] Combining with mathematical equations for explosion or predefined crack path 32

Non-physically based methods Advantages: Intuitive User control Use criterion based… In patterns extracted from images information On the observation Some simplified rules 33

Non-physically based methods Problems Visual quality could be improved Suitable for interactivity application e. g. video games 34

Overview Background Physically-based methods Non-physically based methods Conclusions Future work 35

Conclusions Simulating fracture → a challenging task Modeling the process: Plausability → do physical simplifications Accurate simulation → physical approach Does not exist one ideal model for all kinds of applications 36

Conclusions 37

Problems Quality simulation results small fragments/dust Computation time required real-time to min/hours Trade-off between then Limited user control over animations Physically-based Non-physically methods based methods Visual results Computation time User control 38 Good enough Poor High Low Bit Lot

Methods validation Physically-based methods Compare with experiments on real surfaces Perception [Valette 05; Ramanarayanan 07] Non-physically based methods From a scientific point of view [Lu 07] [Federl 02] 39

Overview Background Physically-based methods Non-physically based methods Conclusions Future work 40

Future work Reproduce a specific pattern simulation Example based techniques restricted to given pattern An open problem Combination of simulation + synthesis [Bosch 11] Extend → cracks and fractures No trivial 41

Challenge Parameters extraction → images/geometry/other Observation/Statistics Find simulation parameters of a specific fracture state of the art study Based on an existing simulation model 42

Inverse model process caracteristics Urban environment Building materials Indirected causes Resistents No elastic 43

Our inverse model process Not accurate model Real time solution Image or acquired models • Extract weathering information Fracture model parameters • Extract parameters using the selected model New fractured scene • Simulate new fractures using the extracted information Simulation image-guided → promess process 44