numerical modelling for phase change materialsarg 4 alice raeli in collaboration with m. azaïez, m....
TRANSCRIPT
![Page 1: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/1.jpg)
Numerical Modelling for Phase Change Materials
Alice RaeliIn collaboration with
M. Azaïez, M. Bergmann and A. Iollo
IMB – INSTITUT DE MATHÉMATIQUES DE BORDEAUX
CANUM 2016 - 9/13 May
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 2: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/2.jpg)
Contents
Abengoa SolarHibrid Material Model
The phase changing materialThe equation
Spatial Discretization - Quadtrees
The Finite Difference Methodarg 4
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 3: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/3.jpg)
Contents
Abengoa SolarHibrid Material Model
The phase changing materialThe equation
Spatial Discretization - Quadtrees
The Finite Difference Methodarg 4
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 4: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/4.jpg)
Contents
Abengoa SolarHibrid Material Model
The phase changing materialThe equation
Spatial Discretization - Quadtrees
The Finite Difference Methodarg 4
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 5: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/5.jpg)
Contents
Abengoa SolarHibrid Material Model
The phase changing materialThe equation
Spatial Discretization - Quadtrees
The Finite Difference Methodarg 4
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 6: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/6.jpg)
Contents
Abengoa SolarHibrid Material Model
The phase changing materialThe equation
Spatial Discretization - Quadtrees
The Finite Difference Methodarg 4
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 7: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/7.jpg)
Contents
Abengoa SolarHibrid Material Model
The phase changing materialThe equation
Spatial Discretization - Quadtrees
The Finite Difference Methodarg 4
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 8: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/8.jpg)
Abengoa Solar - Sevilla, Spain
Renewable and Sustainable Energy : Solar Energy
Heliostats Area, solar power tower and salt tank (ABENGOASolar-Sevilla Spain)
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 9: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/9.jpg)
Hybrid Material
Hybrid Material, Graphite (G) and Salt (S)Graphite (=⇒) Solid Phase
Molten Salt (=⇒) PCM Material (Solid, Liquid)
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 10: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/10.jpg)
Problem Model
Interface Thermal Resistance (=⇒) Jumps on u along the interface
(κ∂nu)S = (κ∂nu)G Flux conservation
R(κ∂nu)G = (uS − uG) = [u] Contact Resistance
R : (Thermal) Contact Resistivity taken constant here.
−div(κ(~x)5 u(~x)) =g(~x), in G ∪ S, (1a)
κ(~x)∂~nu(~x) =0, on ΓN , (1b)
u(~x) =uD(~x), on ΓD, (1c)
[κ(~x)∂~nu(~x)] =0, R(κ∂~nu(~x))S = [u], on γ (1d)
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 11: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/11.jpg)
Spatial Discretization
DefinitionThe term quadtree is used in a more general sense to describe a classof hierarchical data structures whose common property is that they arebased on the principle of recursive decomposition of space. They canbe differentiated on the following bases:
the type of data they are used to represent,
the decomposition process,
the resolution.
We call cell or octant the square/cube that it represents. Each cellmay be the parent of four (eight in 3D) children. The root cell is thebase of the tree and a leaf cell is a cell without any child. The level ofa cell is defined by starting from zero for the root cell and by addingone every time a group of discendant children is added. Each cell Chas two kinds of neighbours: through faces following the axialdirections and through corner following its diagonals directions (edgeconcept is added in 3D).
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 12: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/12.jpg)
To simplify the calculation and focus on certain types of meshes weadd the balancing constraints:
the levels of face neighbouring cells can not differ by more thanone;the levels of diagonally neighbouring can not differ by more thantwo.
Figura: A linear quadtree representation.
The used tool for mesh refinement (PABLOhttp://www.optimad.it/products/pablo/) allows us tobalance the tree working at the same time on non conforming grids.Moreover the used structure is a linear octree, so only the leafs data arestored with evident advantages in memory management.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 13: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/13.jpg)
To localize each cell this tool follows the Z-order and attribute to leafsthe Morton Index. This procedure guarantee at the same time alocalization in the space and a univoque characteristic identificationon the cell itself.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 14: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/14.jpg)
PABLO allows us to manage efficiently these structures and focus onthe method to apply on cells neighbouring configurations. To eachcell is attributed a univoque key that allows us to identify differentconfigurations. We define a function of the level: [L] := L− nL, withL the level of the concerned octant and nL the level of the neighbourso that the value attributed to the key elements are:
0 @ neighbours on this side1 [L] = 02 [L] = −13 [L] = 14 [L] = −25 [L] = 2
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 15: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/15.jpg)
RemarkThe processus to identify each configuration is repeated on threedimensional cases with an appropriate change of the variablesnumber. The dimension of the key has to follow the possibleneighbouring of the cube, so it will be a 26 digit array (at the place of8) and go on for the other dimensions. Also the three dimensionalconfiguration is Z-curve based, the order is shown in figure.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 16: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/16.jpg)
The Finite Difference Method
We consider a cell-centered finite difference method. We present aconfiguration test in figure.
Theorem (Min et al., A supra-convergent finite difference scheme forthe variable coefficient Poisson equation on non-graded grids.)Considering a cell centered finite difference method on c4, if onlyface-adjacent cells are to be used, then there does not exist any locallyconsistent linear scheme on nonuniform Cartesian grids.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 17: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/17.jpg)
We decided to extend the method involving all neighbours and notonly the face adjacent cells. We develop a Taylor expansion for c3with respect to c4, all the others points follow the same reason (ui
stands for the temperature calculated on point i):
u3 = u4 − h∂u4
∂x+ 0 ∗ ∂u4
∂y+
h2
2∂2u4
∂x∂x+ (zero terms) + O(h3)
A Taylor analysis complete on all the neighbours involved, with themrelative linear combinations of the expansions, implies that thecoefficients ai must satisfy the following linear system:
1 1 1 1 1 1 10 −h 0 −h 3h
2 − h2
3h2
0 h h 0 h2 −3h
2 −3h2
0 h2
2 0 h2
29h2
8h2
89h2
80 −h2 0 0 3h2
43h2
4 − 9h2
40 h2
2h2
2 0 h2
89h2
89h2
8
a4a1a2a3a5a6a7
=
000101
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 18: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/18.jpg)
In the example above the concerned points are seven so we candetermine infinite solutions of the complete system but we search aunique one. The idea is to ensure consistency and, at the same time, tominimize the deviation from second-order accuracy. Let M be theconstraints matrix, ~a the weights, ~λ the necessary additionalunknowns,~f the right hand side vector for consistency and F(~a) aweight function. The problem to minimize has the lagrangian form:
L = F(~a)− ~λ(M~a−~f ) (2)
We write minimization problem (2) in matrix form like:
Ax = b⇔
{∂F(~a)∂~a −MT~λ = 0
M~a = ~f
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 19: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/19.jpg)
We call B the submatrix of consistency constraints and C thesubmatrix of order two constrains. Two cases have to be distinguishedfor N the number of concerned points in a configuration:
N ≤ 10 : M = BF(~a) = (1/2)~aT(CTC + αI)~a
Ax =
(((1− α)CTC − αI) −BT
B 0
)x = b
The system satisfies 6 constraints for consistency and minimizes thesecond order constraints and the weights norm.
N > 10 : M =(B
C
)F(~a) = 1/2(~aT~a)
Ax =
(I −
(BC
)T(BC
)0
)x = b
The system satisfies 10 constraints for second order accuracy andminimizes the weights norm.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 20: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/20.jpg)
RemarkWith same reasons presented for the spatial discretization also for themethod the three dimensional case follows the processus modifyingproperly the dimension of the system. We add the z direction, thatmeans other 4 equation for the concistency and 6 for the second order.The two cases presented are so splitted in N ≤ 20 and N > 20.
Figura: A parallel representation of a cube with one jump of level. Study oflocal consistency on the jump.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 21: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/21.jpg)
To check the consistency we reproduced the configuration presentedabove.
Figura: (Left) Error example, a concistency proof with increasing level for alaplacian of a sinusoidal function. (Right) Order study of the method on thetest (logscale).
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 22: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/22.jpg)
Figura: (Left) Alternative error example on the situation presented,concistency proof with increasing level for a laplacian of a sinusoidalfunction. (Right) Order study.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 23: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/23.jpg)
Tests and Results. Penalization
We approached the discontinuity simulation with more kinds ofpenalization tests that confirmed the theorical expectations.
∆u = g +χc
ε(u− u0) (3)
Using u0 constant value we introduce an error on boundaryconditions. We penalized outside and inside the circle.
Figura: (Left) Error example, level of tree equal to 7 (∆x = ∆y = 127 )
(Right)Error order accuracy.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 24: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/24.jpg)
Figura: (a)-(b) Comparison between AMR and uniform mesh for the sameerror obtained at the same depth of the tree. (c)-(d) Error order study in bothcases.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 25: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/25.jpg)
Figura: (Left) Error example with exact boundary condition imposed on thecircle and its (Right) study of order accuracy.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 26: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/26.jpg)
AMR vs Uniform
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 27: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/27.jpg)
We compared the number of points between the mesh refinement onthe penalized part and the uniform mesh. At the same error (saledepth of the tree) the gap becomes relevant in few passages.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 28: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/28.jpg)
Discontinuity with mixed terms
div(κ(~x)5 u(~x)) = f̃ (~x)
u(~x) =18− 1
4∗ ((x− 0.5)2 + (y− 0.5)2))
f̃ (~x) = −κ(~x) + κx(~x)ux(~x) + κy(~x)uy(~y)
Figura: Error (bottom) and numerical solution obtained on AMR refinementfor mixed terms equation.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 29: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/29.jpg)
κ(~x) from 1 to 1.000.000. We choose appropriate constantsA,B, Sharp and we define cell by cellκ(~x) = (A + 1) + B ∗ tanh(Sharp ∗ C(~x)), where C(~x) is anappropriate distance from the point to the level set of discontinuity(circular in the case presented).
Figura: Penalization forced by the κ discontinuity if we impose f (~x) = −1
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 30: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/30.jpg)
Three dimensional
3D sinus uniform
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo
![Page 31: Numerical Modelling for Phase Change Materialsarg 4 Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo Abengoa Solar - Sevilla, Spain Renewable and Sustainable](https://reader034.vdocuments.mx/reader034/viewer/2022050407/5f83fd5ac3057e101749ccda/html5/thumbnails/31.jpg)
Grazie.
Alice Raeli In collaboration with M. Azaïez, M. Bergmann and A. Iollo