Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is...
Transcript of Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is...
![Page 1: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/1.jpg)
Modeling of Thermoelastic Stresses in Thermal Barrier Coatings
Sean P. Donegan and Anthony D. Rollett (PI)UCR/HBCU Review MeetingJune 12, 2013
![Page 2: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/2.jpg)
High Resolution Modeling of Materials for HighTemperature Service
2
University Coal Research: Award Number: DE-FE0003840 Vito Cedro, Program ManagerPROJECT DELIVERABLES:
Identify algorithm for matching both grain and particle microstructuresImplement and deliver software code for synthesis of digital microstructures from experimental images Demonstrate that the Fast Fourier Transform (FFT) code can run as parallel (Message Passing Interface) code on a small clusterDemonstrate that the FFT code can run on a large computer cluster (at least 200 nodes) Characterize a candidate refractory alloy system, build the synthetic microstructure for that alloy from experimental images, perform computer simulations of mechanical response and compare computer simulations with experimental data.Write and submit final report and deliver kinetic database in electronic form suitable for use by other scientists and engineers. Final report will include documentation of the 3D FFT software and the complete code for generating the synthetic microstructures and performing the mechanical response simulations
![Page 3: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/3.jpg)
Outline
Introduction Motivation Objectives
Background Thermoelastic Stress Thermal Barrier Coatings Materials Selection Synthetic Structure Creation
Analytical Techniques Thermoelastic FFT Extreme Value Analysis
Results Resolution Dependence Elastic Energy Density of Thermal Barrier Coatings
MAX Phase Bond Coats Industry Standard Systems
Conclusions
Future Work Microstructure Generation Hot Spots in Relation to Microstructural Features
3
![Page 4: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/4.jpg)
Introduction Metallic components in gas-turbine engines are exposed to extreme levels of temperature. Thermal barrier coatings provide insulation to metallic components in hot gas environments.
4
T.S. Hille et al., Acta Materialia, vol. 57, pp. 2624-2630, 2009
![Page 5: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/5.jpg)
MotivationExperimental assessment of TBC failure involves cycling the component until failure. Scaling prevents FEM simulations from utilizing large structures or limits them to 2D domains.
5
circular TBC test specimen
2D FEM mesh
E.A.G. Shillington and D.R. Clarke, Acta Materialia, vol. 47, pp. 1297-1305, 1999A.M. Karlsson and A.G. Evans, Acta Materialia, vol. 49, pp. 1793-1804, 2001
exposed BC
full thickness top coat
![Page 6: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/6.jpg)
Outline
Background Thermoelastic Stress Thermal Barrier Coatings Materials Selection Synthetic Structure Creation
6
![Page 7: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/7.jpg)
transforming region
isolate region transformation
eigenstrains
surface traction
non-transforming matrix
place back into matrix
eigenstresses
Thermoelastic Stress7
J. D. Eshelby, Proceedings of the Royal Society of London A, vol. 252, pp. 561-569, 1959
![Page 8: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/8.jpg)
transforming region
isolate region transformation
eigenstrains
non-transforming matrix
Thermoelastic Stress8
Eigenstrains resulting from thermal expansion:
strain tensor
thermal expansion tensorapplied temperature
![Page 9: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/9.jpg)
transforming region
isolate region transformation
eigenstrains
non-transforming matrix
Thermoelastic Stress9
Hooke’s Law:
stress tensor stiffness tensorstrain tensor
![Page 10: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/10.jpg)
surface traction place back into matrix
eigenstresses
Thermoelastic Stress10
Total Strain:
Modified Hooke’s Law:
![Page 11: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/11.jpg)
surface traction place back into matrix
eigenstresses
Thermoelastic Stress11
Eigenstresses resulting from thermal expansion:
eigenstress
![Page 12: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/12.jpg)
Thermal Barrier Coatings12
![Page 13: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/13.jpg)
TBC Failure13
![Page 14: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/14.jpg)
Materials Selection14
![Page 15: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/15.jpg)
Synthetic Structure Creation15
Microstructure, especially at the top BC/top coat interface, plays a crucial role in TBC failure. To better appreciate the role of microstructure, DREAM.3D is used to generate test microstructures.
DREAM.3D is a tool used to generate and analyze material microstructure. DREAM.3D can create a 3D microstructure from a set of statistical data. substrate
bond coat
top coat
dream3d.bluequartz.net
![Page 16: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/16.jpg)
Outline
Analytical Techniques Thermoelastic FFT Extreme Value Analysis
16
![Page 17: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/17.jpg)
Fast Fourier Transforms17
The FFT algorithm provides a computationally efficient way to determine discrete direct and indirect Fourier transforms. By re-casting PDEs in frequency space, convolution integrals (Green’s function method) become local (tensor) products. Since all calculations are local save for the FFT, the method has the potential for NlogN scaling to large domain sizes. Full field solutions exists for both the thermoelastic (teFFT) and viscoplastic (vpFFT) cases; both versions have been fully parallelized
![Page 18: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/18.jpg)
Thermoelastic FFT18
We discretize microstructure as a 3D image, or equivalently as a 3D regular grid overlayed on a representative volume element (RVE). Each point/node contains information about the present phase and crystallographic orientation.
The teFFT algorithm computes stress and strain at each node in the grid; no additional mesh is needed.
Due to the FFT, periodic boundary conditions are required, though buffer layers can be used at RVE boundaries
![Page 19: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/19.jpg)
Thermoelastic FFT 1 x C1 x : x * x
2 x x Co : x Co : x x Co : x x Co : x x Co : x x
3 ij,j 0
Cijklo uk,lj x ij,j x 0
periodic boundary conditions in RVE
4 Cijklo Gkm,lj x x im x x 0
5 ij x sym Gik,jl x x kl x d x R3 ij ijkl
o kl
fft ij ijklo kl ij ijkl
o : kl
stiffness tensor of homogeneous solid
perturbation in stress field
NotationStrain: εStress: σStiffness: CPerturbation Stress: τDisplacement: uGreen’s function: GXformed Green’s: Γ
19
R.A. Lebensohn, Acta Materialia, vol. 49, pp. 2723-2737, 2001B.S. Anglin, PhD Thesis, 2012
~ ~
~̂ ^ ^
![Page 20: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/20.jpg)
Extreme Value Analysis 20
Fisher-Tippett-Gnedenko Theorem: The maximum of a random variable converges in distribution to either the Gumbel, Fréchet, or Weibull distribution.
Pickands-Balkema-de Haan Theorem: For some unknown distribution of a random variable, the distribution of values beyond some threshold, u, converges in distribution to the generalized Pareto distribution (GPD).
R.A. Fisher and L.H.C. Tippett, Mathematical Proceedings of the Cambridge Philosophical Society, vol. 24, pp. 180-190, 1927J. Pickands, Annals of Statistics, vol. 3, pp. 119-131, 1975
![Page 21: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/21.jpg)
Extreme Value Analysis21
ξ = shape parameterσ = scale parameterμ = location parameter
x
dens
ity
ξ+
![Page 22: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/22.jpg)
Outline
Results Resolution Dependence Elastic Energy Density of Thermal Barrier Coatings
MAX Phase BCs Industry Standard Systems
22
![Page 23: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/23.jpg)
Resolution Dependence23
323 structure 643 structure
1283 structure The same synthetic microstructure at three resolutions, with equiaxedgrains and random texture.
Material: α-Al2O3∆T = 1000 K
![Page 24: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/24.jpg)
Resolution Dependence24
100s Gpa
323 structure
100s Gpa
643 structure
100s Gpa
1283 structure
![Page 25: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/25.jpg)
MAX Phase BCs25
Mn+1AXn (MAX) phases are ternary hexagonal compounds composed of an early transition metal (M), an A group element (A), and either carbon or nitrogen (X).
![Page 26: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/26.jpg)
MAX Phase BCs26
Nb-16.8 wt% Mo
MAX phase
α-Al2O3
∆T = 1000 K
![Page 27: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/27.jpg)
MAX Phase BCs27
100s Gpa 100s Gpa
Ti2AlC BC V2GeC BC
![Page 28: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/28.jpg)
Industry Standard Systems28
IN718
NiCoCrAlY
columnar YSZ
∆T = 1000 K
![Page 29: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/29.jpg)
Industry Standard Systems29
periodic structure industry standard structure
TBCs used in industry display roughened interfaces either due to deposition technique or cyclic BC creep. A Potts model grain growth approach was used to locally roughen grains near the interfaces.
![Page 30: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/30.jpg)
Industry Standard Systems30
periodic structure industry standard structure
100s Gpa 100s Gpa
![Page 31: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/31.jpg)
Industry Standard Systems31
Histograms of EED thresholded by 0.004 GPa to isolate extreme values. Distributions for structures with flat and rough interfaces display approximately log-normal behavior.
![Page 32: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/32.jpg)
Outline
Conclusions
Future Work Microstructure Generation Hot Spots in Relation to Microstructural Features
32
![Page 33: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/33.jpg)
Future Work33
Microstructure Generation FFT Grid Size: Odd domain sizes exhibit better convergence properties due to the symmetry of the wave function about the zero point of the FFT. FFT Grid Resolution: Resolution analysis indicates that grid resolution has a strong effect on the distribution of extreme field values produced from the teFFT. Higher resolutions are necessary to resolve the TGO, which is up to an order of magnitude thinner than other TBC layers. BC Material Properties: Functionally graded BCs. Top Coat Morphologies: Splat grain morphologies of APS top coats; porosity and cracking in all top coat morphologies.
![Page 34: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/34.jpg)
THANK YOU!
34
![Page 35: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/35.jpg)
Supplemental Slides
35
![Page 36: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/36.jpg)
Thermoelastic Stress
36
![Page 37: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/37.jpg)
Thermal Expansion Tensors37
Transformation of thermal expansion tensor:
Thermal expansion tensors of various crystal symmetries:
![Page 38: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/38.jpg)
Stiffness Tensors38
Symmetry of the stiffness tensor:
Stiffness tensor using Voigt notation:
![Page 39: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/39.jpg)
Stiffness Tensors39
![Page 40: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/40.jpg)
Principal Stresses40
The principal stresses are those stresses normal to planes whose normals are parallel to stress vectors with zero shear component. They are then the eigenvalues of the stress tensor, and the stress tensor can be rewritten:
![Page 41: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/41.jpg)
Thermal Barrier Coatings
41
![Page 42: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/42.jpg)
YSZ Phase Diagrams42
![Page 43: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/43.jpg)
MAX Phase Unit Cell43
![Page 44: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/44.jpg)
MAX Phase Space Group44
![Page 45: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/45.jpg)
Extreme Value Analysis
45
![Page 46: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/46.jpg)
Log-Normal Distribution46
The log-normal distribution describes a random variable whose natural logarithm follows the normal distribution. The cumulative distribution function (cdf) of a log-normal distribution is:
![Page 47: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/47.jpg)
Fisher-Tippett-Gnedenko Theorem47
Consider a set of random variables X1, X2, X3, …, Xn. Let Mn = max(X1, X2, …, Xn). Now suppose two normalizing constants an> 0 and bn exist such that:
According to the first theorem in extreme value theory, referred to as the Fisher-Tippett-Gnedenko theorem, F(x) must be a particular case of the generalized extreme value (GEV) distribution, defined as:
![Page 48: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/48.jpg)
Pickands-Balkema-de Haan Theorem48
Given an unknown underlying distribution function F, the conditional excess distribution, Fu, converges to the GPD as the threshold approaches the right endpoint of F. Alternatively:
The interest is in approximating Fu, the distribution function of X above the threshold u. Fu is defined as follows:
![Page 49: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/49.jpg)
Threshold Choice Plots49
The threshold choice plot shows how an increasing threshold affects the value of the scale (σ) and the shape (ξ) parameter. Consider a random variable X that is distributed as Gξ0,μ0,σ0 . The location parameter, μ0, is the same as the threshold call. Now allow another threshold, μ1 > μ0. The new random variable X | X > μ1 is also described by the GPD. The updated parameters are σ1 = σ0 +ξ0(μ1 −μ0) and ξ1 = ξ0. Let σ∗ = σ1 −ξ1μ1.
In this new parameterization, σ∗ is independent of μ0. Therefore, σ∗ and ξ1 are constant above μ0 if μ0 is a reasonable threshold choice. The threshold choice plot displays graphically σ∗ and ξ1against a range of thresholds. Reasonable threshold choices occur where σ∗ and ξ1 remain constant. Confidence intervals are calculated using the profile likelihood method
![Page 50: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/50.jpg)
Profile Likelihood Confidence Intervals50
Likelihood ratio statistic:
The 95% confidence interval for θ is then all values of θ0 for which the following inequality holds:
![Page 51: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/51.jpg)
Mean Residual Life Plots51
The mean residual life plot consists of the points:
Since the empirical mean is assumed normally distributed by the central limit theorem, confidence intervals can also be plot- ted. The data are a good fit to the GPD where the mean residual life plot follows a straight line.
![Page 52: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/52.jpg)
Return Periods52
Return period related to the probability of non-exceedance:
Quantile function of the GPD:
![Page 53: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/53.jpg)
Thermoelastic FFT
53
![Page 54: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/54.jpg)
Stress Equilibrium54
![Page 55: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/55.jpg)
Compatibility55
For infinitesimal strains, compatibility is satisfied if the following equation holds:
Compatibility ensures a unique strain field is obtainable from a particular displacement field. Conceptually, if a continuous body is thought to be divided into infinitesimal volumes, compatibility describes the necessary conditions under which the body deforms without developing gaps or overlaps between said volumes.
![Page 56: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/56.jpg)
Thermoelastic FFT56
Modified Hooke’s Law with incorporated eigentrains in reference to a homogeneous medium:
Application of stress equilibrium:
![Page 57: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/57.jpg)
Thermoelastic FFT57
Application of Green’s function:
Application of periodic Green’s function to perturbation in stress field:
![Page 58: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/58.jpg)
Thermoelastic FFT58
Application of compatibility:
Application of FFT:
Periodic Green’s function in frequency space:
![Page 59: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/59.jpg)
Augmented Lagrangian59
Non-linear response equation at each iteration:
Non-linear strain field at each iteration:
![Page 60: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/60.jpg)
Initializations for teFFT60
![Page 61: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/61.jpg)
teFFT Algorithm61
1.
2.
3.
4.
5.
6.
7.
![Page 62: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/62.jpg)
teFFT Errors62
Stress field errors:
Strain field errors:
![Page 63: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/63.jpg)
Analysis
63
![Page 64: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/64.jpg)
Stiffness Coefficients64
![Page 65: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/65.jpg)
DREAM3D Statistics65
Sample grain size distribution:
![Page 66: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/66.jpg)
DREAM3D Statistics66
Sample axis ODF pole figures:
![Page 67: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/67.jpg)
Stitching Procedure67
![Page 68: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/68.jpg)
Resolution Dependence: Grain Shape68
1283 structure 643 structure
323 structure
![Page 69: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/69.jpg)
Resolution Dependence: POT Analysis69
1283 structure
![Page 70: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/70.jpg)
Resolution Dependence: POT Analysis70
643 structure
![Page 71: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/71.jpg)
Resolution Dependence: POT Analysis71
323 structure
![Page 72: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/72.jpg)
Resolution Dependence: POT Analysis72
1283 structure
![Page 73: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/73.jpg)
Resolution Dependence: POT Analysis73
643 structure
![Page 74: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/74.jpg)
Resolution Dependence: POT Analysis74
323 structure
![Page 75: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/75.jpg)
MAX Phase BCs: Stress75
Ti2AlC BC Ti2AlN BC
Ti4AlN3 BC V2GeC BC
![Page 76: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/76.jpg)
MAX Phase BCs: Stress76
Nb2AlC BC Ti3AlC2 BC
Ti2SC BC Ti3SiC2 BC
![Page 77: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/77.jpg)
MAX Phase BCs: Stress77
Ti3GeC2 BC V2AlC BC
V2AsC BC
![Page 78: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/78.jpg)
MAX Phase BCs: EED78
Ti2AlC BC Ti2AlN BC
Ti4AlN3 BC V2GeC BC
![Page 79: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/79.jpg)
MAX Phase BCs: EED79
Nb2AlC BC Ti3AlC2 BC
Ti2SC BC Ti3SiC2 BC
![Page 80: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/80.jpg)
MAX Phase BCs: EED80
Ti3GeC2 BC V2AlC BC
V2AsC BC
![Page 81: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/81.jpg)
MAX Phase BCs: Summary81
![Page 82: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/82.jpg)
MAX Phase BCs: Summary82
![Page 83: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/83.jpg)
Industry Standard Systems: Stress83
Periodic, flat Periodic, rough
Industry, flat Industry, h
![Page 84: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/84.jpg)
Industry Standard Systems: Stress84
Periodic, flat Periodic, rough
Industry, flat Industry, h
![Page 85: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/85.jpg)
Industry Standard Systems: POT Analysis85
Industry, flat
![Page 86: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/86.jpg)
Industry Standard Systems: POT Analysis86
Industry, h
![Page 87: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/87.jpg)
Industry Standard Systems: POT Analysis87
Industry, flat
![Page 88: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/88.jpg)
Industry Standard Systems: POT Analysis88
Industry, h
![Page 89: Modeling of Thermoelastic Rollett Donegan · 2013-10-07 · The new random variable X | X > μ 1 is also described by the GPD. The updated parameters are σ 1 = σ 0 +ξ 0(μ 1 −μ](https://reader034.fdocuments.net/reader034/viewer/2022050517/5fa0d20ec1743121654093d0/html5/thumbnails/89.jpg)
“Harpy’s Eagle, world’s most vicious raptor…”
89