Data-Driven Filtered Reduced Order Models of Nonlinear Systems › data › pages ›...

63
Data-Driven Filtered Reduced Order Models of Nonlinear Systems T. Iliescu Department of Mathematics ICAM Virginia Tech MoRePaS 2018 Nantes April 11, 2018 National Science Foundation DMS-1522656 2015-2018 T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 1 / 17

Transcript of Data-Driven Filtered Reduced Order Models of Nonlinear Systems › data › pages ›...

  • Data-Driven FilteredReduced Order Modelsof Nonlinear Systems

    T. Iliescu

    Department of MathematicsICAM

    Virginia Tech

    MoRePaS 2018

    NantesApril 11, 2018

    National Science Foundation DMS-1522656 2015-2018

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 1 / 17

  • Outline

    1 Problem

    2 Solution

    3 Connections

    4 Conclusions

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 2 / 17

  • Outline

    Collaborators

    Muhammad Mohebujjaman (postdoc, Virginia Tech)

    Xuping Xie (postdoc, Oak Ridge National Laboratory)

    Leo Rebholz (professor, Clemson University)

    Birgul Koc (student, Virginia Tech)

    Changhong Mou (student, Virginia Tech)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 3 / 17

  • Outline

    Collaborators

    Muhammad Mohebujjaman (postdoc, Virginia Tech)

    Xuping Xie (postdoc, Oak Ridge National Laboratory)

    Leo Rebholz (professor, Clemson University)

    Birgul Koc (student, Virginia Tech)

    Changhong Mou (student, Virginia Tech)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 3 / 17

  • Outline

    Collaborators

    Muhammad Mohebujjaman (postdoc, Virginia Tech)

    Xuping Xie (postdoc, Oak Ridge National Laboratory)

    Leo Rebholz (professor, Clemson University)

    Birgul Koc (student, Virginia Tech)

    Changhong Mou (student, Virginia Tech)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 3 / 17

  • Outline

    Collaborators

    Muhammad Mohebujjaman (postdoc, Virginia Tech)

    Xuping Xie (postdoc, Oak Ridge National Laboratory)

    Leo Rebholz (professor, Clemson University)

    Birgul Koc (student, Virginia Tech)

    Changhong Mou (student, Virginia Tech)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 3 / 17

  • Outline

    Collaborators

    Muhammad Mohebujjaman (postdoc, Virginia Tech)

    Xuping Xie (postdoc, Oak Ridge National Laboratory)

    Leo Rebholz (professor, Clemson University)

    Birgul Koc (student, Virginia Tech)

    Changhong Mou (student, Virginia Tech)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 3 / 17

  • Problem

    General Problem

    POD

    RBM

    DMD

    systems theoretic ROMs (Serkan ,)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 4 / 17

  • Problem

    General Problem

    POD

    RBM

    DMD

    systems theoretic ROMs (Serkan ,)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 4 / 17

  • Problem

    General Problem

    POD

    RBM

    DMD

    systems theoretic ROMs (Serkan ,)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 4 / 17

  • Problem

    General Problem

    POD

    RBM

    DMD

    systems theoretic ROMs (Serkan ,)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 4 / 17

  • Problem

    What is the Problem?

    (I) PDE•u = f(u) given

    (II) FEM(

    •uh ,vh

    )=

    (f(uh) ,vh

    )given

    (III) ROM

    r given

    {ϕ1, . . . ,ϕr} given

    model ???

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 5 / 17

  • Problem

    What is the Problem?

    (I) PDE•u = f(u) given

    (II) FEM(

    •uh ,vh

    )=

    (f(uh) ,vh

    )given

    (III) ROM

    r given

    {ϕ1, . . . ,ϕr} given

    model ???

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 5 / 17

  • Problem

    What is the Problem?

    (I) PDE•u = f(u) given

    (II) FEM(

    •uh ,vh

    )=

    (f(uh) ,vh

    )given

    (III) ROM

    r given

    {ϕ1, . . . ,ϕr} given

    model ???

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 5 / 17

  • Problem

    What is the Problem?

    (I) PDE•u = f(u) given

    (II) FEM(

    •uh ,vh

    )=

    (f(uh) ,vh

    )given

    (III) ROM

    r given

    {ϕ1, . . . ,ϕr} given

    model ???

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 5 / 17

  • Problem

    What is the Problem?

    (I) PDE•u = f(u) given

    (II) FEM(

    •uh ,vh

    )=

    (f(uh) ,vh

    )given

    (III) ROM

    r given

    {ϕ1, . . . ,ϕr} given

    model ???

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 5 / 17

  • Problem

    What is the Problem?

    (I) PDE•u = f(u) given

    (II) FEM(

    •uh ,vh

    )=

    (f(uh) ,vh

    )given

    (III) ROM

    r given

    {ϕ1, . . . ,ϕr} given

    model ???

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 5 / 17

  • Solution

    Standard Solution Discretization Levels

    (I) PDE dim =∞

    (II) FEM dim = N = O(106)

    (III) ROM dim = r = O(10)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 6 / 17

  • Solution

    Standard Solution Discretization Levels

    (I) PDE dim =∞

    (II) FEM dim = N = O(106)

    (III) ROM dim = r = O(10)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 6 / 17

  • Solution

    Standard Solution Discretization Levels

    (I) PDE dim =∞

    (II) FEM dim = N = O(106)

    (III) ROM dim = r = O(10)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 6 / 17

  • Solution

    Standard Solution Model

    ur =r∑

    j=1

    aj ϕj

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )i = 1, . . . , r

    Galerkin ROM 1

    1NSE•a = A a + a> B a

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 7 / 17

  • Solution

    Standard Solution Model

    ur =r∑

    j=1

    aj ϕj

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )i = 1, . . . , r

    Galerkin ROM 1

    1NSE•a = A a + a> B a

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 7 / 17

  • Solution

    Standard Solution Model

    ur =r∑

    j=1

    aj ϕj

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )i = 1, . . . , r

    Galerkin ROM 1

    1NSE•a = A a + a> B a

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 7 / 17

  • Solution

    New Solution Discretization Levels

    (I) PDE dim =∞

    (II) FEM dim = N = O(106)

    (III) FEM-ROM dim = d = O(103)

    (IV) ROM dim = r = O(10)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 8 / 17

  • Solution

    New Solution Discretization Levels

    (I) PDE dim =∞

    (II) FEM dim = N = O(106)

    (III) FEM-ROM dim = d = O(103)

    (IV) ROM dim = r = O(10)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 8 / 17

  • Solution

    New Solution Discretization Levels

    (I) PDE dim =∞

    (II) FEM dim = N = O(106)

    (III) FEM-ROM dim = d = O(103)

    (IV) ROM dim = r = O(10)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 8 / 17

  • Solution

    New Solution Discretization Levels

    (I) PDE dim =∞

    (II) FEM dim = N = O(106)

    (III) FEM-ROM dim = d = O(103)

    (IV) ROM dim = r = O(10)

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 8 / 17

  • Solution

    New Solution Model

    ud =d∑

    j=1

    aj ϕj

    (•

    ud ,ϕi

    )=

    (f(ud) ,ϕi

    )i = 1, . . . , r

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )+

    [(f(ud) ,ϕi

    )−(

    f(ur) ,ϕi

    )]i = 1, . . . , r

    Galerkin ROM + correction

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 9 / 17

  • Solution

    New Solution Model

    ud =d∑

    j=1

    aj ϕj

    (•

    ud ,ϕi

    )=

    (f(ud) ,ϕi

    )i = 1, . . . , r

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )+

    [(f(ud) ,ϕi

    )−(

    f(ur) ,ϕi

    )]i = 1, . . . , r

    Galerkin ROM + correction

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 9 / 17

  • Solution

    New Solution Model

    ud =d∑

    j=1

    aj ϕj

    (•

    ud ,ϕi

    )=

    (f(ud) ,ϕi

    )i = 1, . . . , r

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )+

    [(f(ud) ,ϕi

    )−(

    f(ur) ,ϕi

    )]i = 1, . . . , r

    Galerkin ROM + correction

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 9 / 17

  • Solution

    New Solution Model

    ud =d∑

    j=1

    aj ϕj

    (•

    ud ,ϕi

    )=

    (f(ud) ,ϕi

    )i = 1, . . . , r

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )+

    [(f(ud) ,ϕi

    )−(

    f(ur) ,ϕi

    )]i = 1, . . . , r

    Galerkin ROM + correction

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 9 / 17

  • Solution

    New Solution Data-Driven Model

    [(f(ud) ,ϕi

    )−(

    f(ur) ,ϕi

    )]data-driven≈

    (g(ur) ,ϕi

    )

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )+

    (g(ur) ,ϕi

    )i = 1, . . . , r

    Data-Driven Filtered ROM (DDF-ROM) 2

    2NSE•a =

    (A + Ã

    )a + a>

    (B + B̃

    )a

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 10 / 17

  • Solution

    New Solution Data-Driven Model

    [(f(ud) ,ϕi

    )−(

    f(ur) ,ϕi

    )]data-driven≈

    (g(ur) ,ϕi

    )

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )+

    (g(ur) ,ϕi

    )i = 1, . . . , r

    Data-Driven Filtered ROM (DDF-ROM) 2

    2NSE•a =

    (A + Ã

    )a + a>

    (B + B̃

    )a

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 10 / 17

  • Solution

    New Solution Data-Driven Model

    [(f(ud) ,ϕi

    )−(

    f(ur) ,ϕi

    )]data-driven≈

    (g(ur) ,ϕi

    )

    (•

    ur ,ϕi

    )=

    (f(ur) ,ϕi

    )+

    (g(ur) ,ϕi

    )i = 1, . . . , r

    Data-Driven Filtered ROM (DDF-ROM) 2

    2NSE•a =

    (A + Ã

    )a + a>

    (B + B̃

    )a

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 10 / 17

  • Solution

    DDF-ROM Reproductive

    Re = 1000

    0.5 1 1.5 2 2.5 3

    t

    0.605

    0.61

    0.615

    0.62

    En

    erg

    y

    r=4

    0.5 1 1.5 2 2.5 3

    t

    2.5

    3

    3.5

    4

    4.5

    Dra

    g

    N=4

    0.5 1 1.5 2 2.5 3

    t

    -3

    -2

    -1

    0

    1

    2

    3

    Lift

    N=4

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 11 / 17

  • Solution

    DDF-ROM + Constraints Predictive

    Re = 1000

    50% data

    0.5 1 1.5 2 2.5 3 3.5 4 4.5 5

    t

    0.599

    0.6

    0.601

    0.602

    0.603

    0.604

    0.605

    0.606

    En

    erg

    y

    r=3

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 12 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Correction = Numerical Dissipation

    eddy viscosity ROMs

    mixing length: Lumley et al

    Smagorinsky: Noack et al, Ullman & Lang, Wang et al, Rozza et al

    variational multiscale: Iollo et al, Wang et al , . . .

    dynamic subgrid-scale: Wang et al

    numerical stabilizationNoack et al, Iollo et al, Karniadakis et al, Farhat et al, Amsallem et al, Carlberg et al,Kalashnikova et al, Balajewicz et al, Wells et al,. . .

    Schneier Friday talk

    Stabile Thursday talk

    Ali, Hijazi Wednesday poster

    Taddei Wednesday poster

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 13 / 17

  • Connections

    Galerkin ROM Stable, Not Accurate

    stable

    ‖unr ‖2 + ν ∆tM∑

    n=1

    ‖∇unr ‖2 ≤ data

    not accurate

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 14 / 17

  • Connections

    Galerkin ROM Stable, Not Accurate

    stable

    ‖unr ‖2 + ν ∆tM∑

    n=1

    ‖∇unr ‖2 ≤ data

    not accurate

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 14 / 17

  • Connections

    Galerkin ROM Stable, Not Accurate

    stable

    ‖unr ‖2 + ν ∆tM∑

    n=1

    ‖∇unr ‖2 ≤ data

    not accurate

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 14 / 17

  • Connections

    Connections

    Large Eddy Simulationu u

    h

    Variational Multiscale

    Nonlinear Galerkin

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 15 / 17

  • Connections

    Connections

    Large Eddy Simulationu u

    h

    Variational Multiscale

    Nonlinear Galerkin

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 15 / 17

  • Connections

    Connections

    Large Eddy Simulationu u

    h

    Variational Multiscale

    Nonlinear Galerkin

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 15 / 17

  • Connections

    Related Work

    no ROM

    20th century Adrian, Moser

    21st century Iaccarino, Duraisamy, Ling, Xiao

    ROM

    21st century Chorin, Duraisamy, Codina, Majda, San, Liu

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 16 / 17

  • Connections

    Related Work

    no ROM

    20th century Adrian, Moser

    21st century Iaccarino, Duraisamy, Ling, Xiao

    ROM

    21st century Chorin, Duraisamy, Codina, Majda, San, Liu

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 16 / 17

  • Connections

    Related Work

    no ROM

    20th century Adrian, Moser

    21st century Iaccarino, Duraisamy, Ling, Xiao

    ROM

    21st century Chorin, Duraisamy, Codina, Majda, San, Liu

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 16 / 17

  • Connections

    Related Work

    no ROM

    20th century Adrian, Moser

    21st century Iaccarino, Duraisamy, Ling, Xiao

    ROM

    21st century Chorin, Duraisamy, Codina, Majda, San, Liu

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 16 / 17

  • Connections

    Related Work

    no ROM

    20th century Adrian, Moser

    21st century Iaccarino, Duraisamy, Ling, Xiao

    ROM

    21st century Chorin, Duraisamy, Codina, Majda, San, Liu

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 16 / 17

  • Conclusions

    Conclusions

    Galerkin ROM + correction

    correction data-driven

    Re = 100, Re = 500, Re = 1000

    reproductive X

    predictive (cross-validation) X

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 17 / 17

  • Conclusions

    Conclusions

    Galerkin ROM + correction

    correction data-driven

    Re = 100, Re = 500, Re = 1000

    reproductive X

    predictive (cross-validation) X

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 17 / 17

  • Conclusions

    Conclusions

    Galerkin ROM + correction

    correction data-driven

    Re = 100, Re = 500, Re = 1000

    reproductive X

    predictive (cross-validation) X

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 17 / 17

  • Conclusions

    Conclusions

    Galerkin ROM + correction

    correction data-driven

    Re = 100, Re = 500, Re = 1000

    reproductive X

    predictive (cross-validation) X

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 17 / 17

  • Conclusions

    Conclusions

    Galerkin ROM + correction

    correction data-driven

    Re = 100, Re = 500, Re = 1000

    reproductive X

    predictive (cross-validation) X

    T. Iliescu (Mathematics) DDF-ROM MoRePaS 2018 17 / 17

    ProblemSolutionConnectionsConclusions