New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

download New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

of 95

Transcript of New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    1/95

    New mathematical models for suspension bridges

    Filippo Gazzola

    Dipartimento di Matematica - Politecnico di Milano

    Alghero - June 2013

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    2/95

    FIRST SUSPENSION BRIDGES

    The first pictures of suspension bridges appear in  Machinae Novae 

    (1615) by  Fausto Veranzio  from the Republic of Venice.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    3/95

    FIRST SUSPENSION BRIDGES

    The first pictures of suspension bridges appear in  Machinae Novae 

    (1615) by  Fausto Veranzio  from the Republic of Venice.

    However, these bridges were never built.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    4/95

    FIRST SUSPENSION BRIDGES

    The first pictures of suspension bridges appear in  Machinae Novae 

    (1615) by  Fausto Veranzio  from the Republic of Venice.

    However, these bridges were never built.

    The first suspension bridges were built only about two centurieslater. Many bridges had serious problems under the action of thewind or of traffic loads.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    5/95

    The most impressive failure of history is certainly the TacomaNarrows Bridge collapse in November 1940.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    6/95

    The most impressive failure of history is certainly the TacomaNarrows Bridge collapse in November 1940.

    Collapsed and current Tacoma Narrows Bridge:

    The Official Report•  O.H. Ammann, T. von Kármán, G.B. Woodruff,  The failure of   

    the Tacoma Narrows Bridge , Federal Works Agency (1941)considers ...the crucial event in the collapse to be the suddenchange from a vertical to a torsional mode of oscillation.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    7/95

    The most impressive failure of history is certainly the TacomaNarrows Bridge collapse in November 1940.

    Collapsed and current Tacoma Narrows Bridge:

    The Official Report•  O.H. Ammann, T. von Kármán, G.B. Woodruff,  The failure of   

    the Tacoma Narrows Bridge , Federal Works Agency (1941)considers ...the crucial event in the collapse to be the suddenchange from a vertical to a torsional mode of oscillation.

    This leads to the following main question:   why did torsional

    oscillations appear suddenly?F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    8/95

    There have been many attempts to answer:

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    9/95

    There have been many attempts to answer:

    - resonance due to the frequency of the wind

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    10/95

    There have been many attempts to answer:

    - resonance due to the frequency of the wind- mistake in the project, structural failure

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    11/95

    There have been many attempts to answer:

    - resonance due to the frequency of the wind- mistake in the project, structural failure

    - flutter theory

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    12/95

    There have been many attempts to answer:

    - resonance due to the frequency of the wind- mistake in the project, structural failure

    - flutter theory

    - vortex shedding, von Kármán vortices

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    13/95

    There have been many attempts to answer:

    - resonance due to the frequency of the wind- mistake in the project, structural failure

    - flutter theory

    - vortex shedding, von Kármán vortices

    - angle of attack of the wind

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    14/95

    There have been many attempts to answer:

    - resonance due to the frequency of the wind- mistake in the project, structural failure

    - flutter theory

    - vortex shedding, von Kármán vortices

    - angle of attack of the wind- ...

    None of these explanations gives a full answer to the mainquestion:

    why did the longitudinal oscillation transform suddenly into atorsional oscillation?

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    15/95

    Since then, many bridges have stiffening trusses which are veryrigid, heavy and... expensive!

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    16/95

    Since then, many bridges have stiffening trusses which are veryrigid, heavy and... expensive!

    The problem has been solved, but... is it understood?

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    17/95

    Since then, many bridges have stiffening trusses which are veryrigid, heavy and... expensive!

    The problem has been solved, but... is it understood?

    There is a lack of  reliable mathematical models  which should

    also display torsional oscillations.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    18/95

    Since then, many bridges have stiffening trusses which are veryrigid, heavy and... expensive!

    The problem has been solved, but... is it understood?

    There is a lack of  reliable mathematical models  which should

    also display torsional oscillations.

    In particular,  the bridge should not be modeled as a beam.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    The cross section of the bridge may be modeled by

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    19/95

    The  cross section  of the bridge may be modeled by

    •  Y. Rocard,  Dynamic instability: automobiles, aircraft, suspensionbridges , 1957

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    The cross section of the bridge may be modeled by

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    20/95

    The  cross section  of the bridge may be modeled by

    •  Y. Rocard,  Dynamic instability: automobiles, aircraft, suspensionbridges , 1957

    The  equations  describing this double oscillator are

    m2

    3 θ̈ =  cos θ

    f   (y  − sin θ) − f   (y  +  sin θ)

    m ÿ  = −

    f   (y  − sin θ) + f   (y  +  sin θ)

    .

    •  J. McKenna, American Mathematical Monthly, 1999F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    21/95

    The force   f   should take into account both the restoring elasticforce due to the hanger and gravity: McKenna takes   f    piecewiselinear, that is, linear unless the hanger slackens. He then adds

    periodic forcing terms and damping.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    22/95

    The force   f   should take into account both the restoring elasticforce due to the hanger and gravity: McKenna takes   f    piecewiselinear, that is, linear unless the hanger slackens. He then adds

    periodic forcing terms and damping.He is so able to numerically replicate   the sudden transition fromvertical oscillations to torsional oscillations, but he writes that:...the range of parameters over which the transition from vertical 

    to torsional motion was observed was physically unreasonable ...

    the restoring force due to the cables was oversimplified ... it was 

    necessary to impose small torsional forcing...

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    23/95

    The force   f   should take into account both the restoring elasticforce due to the hanger and gravity: McKenna takes   f    piecewiselinear, that is, linear unless the hanger slackens. He then adds

    periodic forcing terms and damping.He is so able to numerically replicate   the sudden transition fromvertical oscillations to torsional oscillations, but he writes that:...the range of parameters over which the transition from vertical 

    to torsional motion was observed was physically unreasonable ...

    the restoring force due to the cables was oversimplified ... it was 

    necessary to impose small torsional forcing...

    We may conclude that

    the McKenna model is quite promising, for the first time the

    sudden transition between oscillations is visible.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    24/95

    The force   f   should take into account both the restoring elasticforce due to the hanger and gravity: McKenna takes   f    piecewiselinear, that is, linear unless the hanger slackens. He then adds

    periodic forcing terms and damping.He is so able to numerically replicate   the sudden transition fromvertical oscillations to torsional oscillations, but he writes that:...the range of parameters over which the transition from vertical 

    to torsional motion was observed was physically unreasonable ...

    the restoring force due to the cables was oversimplified ... it was 

    necessary to impose small torsional forcing...

    We may conclude that

    the McKenna model is quite promising, for the first time the

    sudden transition between oscillations is visible.

    But it may be improved... some witnesses claim that the hangerswere never slacken at the Tacoma Bridge.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    25/95

    Together with Gianni Arioli (PoliMi) we have  stripped  the modelby dropping both internal frictions (damping) and external sources(forcing).

    The system is then   isolated and conserves energy.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    26/95

    Together with Gianni Arioli (PoliMi) we have  stripped  the modelby dropping both internal frictions (damping) and external sources(forcing).

    The system is then   isolated and conserves energy.

    We assume that the moving rod is linked to two fixed rods throughan elastic membrane with elastic constant  K  > 0:

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We considered nonlinear forces also for small elongations of the

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    27/95

    hangers:   f   (s ) = s  + εφ(s ) with  φ(s ) = o (s ) as  s  → 0. The valueof  ε > 0 and the form of  φ  play  no qualitative role.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We considered nonlinear forces also for small elongations of the

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    28/95

    hangers:   f   (s ) = s  + εφ(s ) with  φ(s ) = o (s ) as  s  → 0. The valueof  ε > 0 and the form of  φ  play  no qualitative role.

    Basic requirements:   f   (0) = 0,   f   (s ) > 0.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We considered nonlinear forces also for small elongations of the

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    29/95

    hangers:   f   (s ) = s  + εφ(s ) with  φ(s ) = o (s ) as  s  → 0. The valueof  ε > 0 and the form of  φ  play  no qualitative role.

    Basic requirements:   f   (0) = 0,   f   (s ) > 0. Let us take

    f   (s ) = s  + s 2 + s 3

    and  m =   =  K  = 1. Set  F (s ) = s 0   f   (σ)d σ  and

    U (θ, y ) = F (y  + sin θ) + F (y  − sinθ) + y 2 + θ2 .

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We considered nonlinear forces also for small elongations of the( ) ( ) ( ) ( )

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    30/95

    hangers:   f   (s ) = s  + εφ(s ) with  φ(s ) = o (s ) as  s  → 0. The valueof  ε > 0 and the form of  φ  play  no qualitative role.

    Basic requirements:   f   (0) = 0,   f   (s ) > 0. Let us take

    f   (s ) = s  + s 2 + s 3

    and  m =   =  K  = 1. Set  F (s ) = s 0   f   (σ)d σ  and

    U (θ, y ) = F (y  + sin θ) + F (y  − sinθ) + y 2 + θ2 .

    Then the equations becomeθ̈

    3 + U θ(θ, y ) = 0 ,   ÿ  + U y (θ, y ) = 0

    which have a (conserved) energy given by

    E  =θ̇2

    6  +

     ẏ 2

    2  + U (θ, y ) .

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We considered nonlinear forces also for small elongations of theh f ( ) φ( ) i h φ( ) ( ) 0 Th l

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    31/95

    hangers:   f   (s ) = s  + εφ(s ) with  φ(s ) = o (s ) as  s  → 0. The valueof  ε > 0 and the form of  φ  play  no qualitative role.

    Basic requirements:   f   (0) = 0,   f   (s ) > 0. Let us take

    f   (s ) = s  + s 2 + s 3

    and  m =   =  K  = 1. Set  F (s ) = s 0   f   (σ)d σ  and

    U (θ, y ) = F (y  + sin θ) + F (y  − sinθ) + y 2 + θ2 .

    Then the equations becomeθ̈

    3 + U θ(θ, y ) = 0 ,   ÿ  + U y (θ, y ) = 0

    which have a (conserved) energy given by

    E  =θ̇2

    6  +

     ẏ 2

    2  + U (θ, y ) .

    Finally, we consider initial data such as

    (y (0),  ẏ (0)) = (y 0, y 1) ,   0 < |θ(0)| + |θ̇(0)|

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    32/95

    (y (0), ẏ (0)) = (0, 2.8) ⇒ E  = 3.92

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    33/95

    (y (0), ẏ (0)) = (0, 2.8) ⇒ E  = 3.92

    (y (0), ẏ (0)) = (0, 3) ⇒ E  = 4.5

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    34/95

    (y (0), ẏ (0)) = (0, 2.8) ⇒ E  = 3.92

    (y (0), ẏ (0)) = (0, 3) ⇒ E  = 4.5

    (y (0), ẏ (0)) = (0, 3.2) ⇒ E  = 5.12

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    35/95

    (y (0), ẏ (0)) = (0, 2.8) ⇒ E  = 3.92

    (y (0), ẏ (0)) = (0, 3) ⇒ E  = 4.5

    (y (0), ẏ (0)) = (0, 3.2) ⇒ E  = 5.12There is a  sudden transfer of oscillations which becomes morerelevant for increasing energy; no transfer is visible if  E   3.56.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    ENERGY TRANSFER  (only for large energies).

    E E ( ) E ( ) E ( )

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    36/95

    We decompose  E T   = E y (t ) + E θ(t ) + E y θ(t ).

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    ENERGY TRANSFER  (only for large energies).

    W d E E ( ) E ( ) E ( )

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    37/95

    We decompose  E T   = E y (t ) + E θ(t ) + E y θ(t ).

    (y (0), ẏ (0)) = (0, 2.8) ⇒ E  = 3.92

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    ENERGY TRANSFER  (only for large energies).

    W d E E ( ) + E ( ) + E ( )

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    38/95

    We decompose  E T   = E y (t ) + E θ(t ) + E y θ(t ).

    (y (0), ẏ (0)) = (0, 2.8) ⇒ E  = 3.92

    (y (0), ẏ (0)) = (0, 3) ⇒ E  = 4.5

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    ENERGY TRANSFER  (only for large energies).

    W d E E (t) + E (t) + E (t)

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    39/95

    We decompose  E T   = E y (t ) + E θ(t ) + E y θ(t ).

    (y (0), ẏ (0)) = (0, 2.8) ⇒ E  = 3.92

    (y (0), ẏ (0)) = (0, 3) ⇒ E  = 4.5

    (y (0), ẏ (0)) = (0, 3.2) ⇒ E  = 5.12F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Can we explain this fact?

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    40/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Can we explain this fact?  The phase space is 4D but, due to theconservation of energy, the dynamics takes place in a 3D manifold.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    41/95

    conservation of energy, the dynamics takes place in a 3D manifold.We take a  Poincaré section: we represent the (θ̇, θ)-intersectionsof the solutions of the system with the plane  y  = 0 when ẏ  > 0 at

    energies  E  = 3.4,   3.5,   3.6,   3.8.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Can we explain this fact?  The phase space is 4D but, due to theconservation of energy, the dynamics takes place in a 3D manifold.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    42/95

    gy, y pWe take a  Poincaré section: we represent the (θ̇, θ)-intersectionsof the solutions of the system with the plane  y  = 0 when ẏ  > 0 at

    energies  E  = 3.4,   3.5,   3.6,   3.8.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Can we explain this fact?  The phase space is 4D but, due to theconservation of energy, the dynamics takes place in a 3D manifold.

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    43/95

    gy, y pWe take a  Poincaré section: we represent the (θ̇, θ)-intersectionsof the solutions of the system with the plane  y  = 0 when ẏ  > 0 at

    energies  E  = 3.4,   3.5,   3.6,   3.8.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Can we explain this fact?  The phase space is 4D but, due to theconservation of energy, the dynamics takes place in a 3D manifold.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    44/95

    gy y pWe take a  Poincaré section: we represent the (θ̇, θ)-intersectionsof the solutions of the system with the plane  y  = 0 when ẏ  > 0 atenergies

     E  = 3.4,   3.5,   3.6,   3.8.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Can we explain this fact?  The phase space is 4D but, due to theconservation of energy, the dynamics takes place in a 3D manifold.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    45/95

    We take a  Poincaré section: we represent the (θ̇, θ)-intersectionsof the solutions of the system with the plane  y  = 0 when ẏ  > 0 atenergies  E  = 3.4,   3.5,   3.6,   3.8.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Can we explain this fact?  The phase space is 4D but, due to theconservation of energy, the dynamics takes place in a 3D manifold.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    46/95

    We take a  Poincaré section: we represent the (θ̇, θ)-intersectionsof the solutions of the system with the plane  y  = 0 when ẏ  > 0 atenergies  E  = 3.4,   3.5,   3.6,   3.8.

    The origin is a  stable  fixed point for the  Poincaré map  in the toppictures, whereas it is  unstable   in the bottom pictures.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    47/95

    Since the system is conservative, det(JP E (0, 0)) = 1, so eitherboth its eigenvalues have modulus 1, or they are both real.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    48/95

    Since the system is conservative, det(JP E (0, 0)) = 1, so eitherboth its eigenvalues have modulus 1, or they are both real.

    •   if  |λ1| = |λ2| = 1 and  λ2  = λ1  the system is   torsionally stable;•   if  λ1, λ2 ∈ R  and  |λ1|

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    49/95

    Since the system is conservative, det(JP E (0, 0)) = 1, so eitherboth its eigenvalues have modulus 1, or they are both real.

    •   if  |λ1| = |λ2| = 1 and  λ2  = λ1  the system is   torsionally stable;•   if  λ1, λ2 ∈ R  and  |λ1|  0 such that the system is torsionally stable

    whenever 0 < E  

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    50/95

    Since the system is conservative, det(JP E (0, 0)) = 1, so eitherboth its eigenvalues have modulus 1, or they are both real.

    •   if  |λ1| = |λ2| = 1 and  λ2  = λ1  the system is   torsionally stable;•   if  λ1, λ2 ∈ R  and  |λ1|  0 such that the system is torsionally stable

    whenever 0 < E  

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    51/95

    Since the system is conservative, det(JP E (0, 0)) = 1, so eitherboth its eigenvalues have modulus 1, or they are both real.

    •   if  |λ1| = |λ2| = 1 and  λ2  = λ1  the system is   torsionally stable;•   if  λ1, λ2 ∈ R  and  |λ1|  0 such that the system is torsionally stable

    whenever 0 < E  

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    52/95

    1) Why did we first exclude resonance and now we speak againabout resonance?

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    NATURAL QUESTIONS:

    1) Wh did fi l d d k i

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    53/95

    1) Why did we first exclude resonance and now we speak againabout resonance?One must distinguish between external resonance and

    internal resonance.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    NATURAL QUESTIONS:

    1) Wh did fi l d d k i

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    54/95

    1) Why did we first exclude resonance and now we speak againabout resonance?One must distinguish between external resonance and

    internal resonance.

    2) What happens if we change the parameters in the model?

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    NATURAL QUESTIONS:

    1) Wh did fi t l d d k i

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    55/95

    1) Why did we first exclude resonance and now we speak againabout resonance?One must distinguish between external resonance and

    internal resonance.

    2) What happens if we change the parameters in the model?Quantitative variations of the critical energy threshold, the

    kind of bifurcation may be different. The model is robust.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    NATURAL QUESTIONS:

    1) Wh did fi t l d d k i

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    56/95

    1) Why did we first exclude resonance and now we speak againabout resonance?One must distinguish between external resonance and

    internal resonance.

    2) What happens if we change the parameters in the model?Quantitative variations of the critical energy threshold, the

    kind of bifurcation may be different. The model is robust.

    3) What happens if we add damping and forcing terms?

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    NATURAL QUESTIONS:

    1) Why did we first exclude resonance and now we speak again

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    57/95

    1) Why did we first exclude resonance and now we speak againabout resonance?One must distinguish between external resonance and

    internal resonance.

    2) What happens if we change the parameters in the model?Quantitative variations of the critical energy threshold, the

    kind of bifurcation may be different. The model is robust.

    3) What happens if we add damping and forcing terms?Only quantitative variations of the width of the oscillations

    and of the critical energy threshold.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    NATURAL QUESTIONS:

    1) Why did we first exclude resonance and now we speak again

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    58/95

    1) Why did we first exclude resonance and now we speak againabout resonance?One must distinguish between external resonance and

    internal resonance.

    2) What happens if we change the parameters in the model?Quantitative variations of the critical energy threshold, the

    kind of bifurcation may be different. The model is robust.

    3) What happens if we add damping and forcing terms?Only quantitative variations of the width of the oscillations

    and of the critical energy threshold.

    50 100 150 200 250 300 350

    0.6

    0.4

    0.2

    0.2

    0.4

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    NATURAL QUESTIONS:

    1) Why did we first exclude resonance and now we speak again

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    59/95

    1) Why did we first exclude resonance and now we speak againabout resonance?One must distinguish between external resonance and

    internal resonance.

    2) What happens if we change the parameters in the model?Quantitative variations of the critical energy threshold, the

    kind of bifurcation may be different. The model is robust.

    3) What happens if we add damping and forcing terms?Only quantitative variations of the width of the oscillations

    and of the critical energy threshold.

    50 100 150 200 250 300 350

    0.6

    0.4

    0.2

    0.2

    0.4

    50 100 150 200 250 300 350

    0.4

    0.2

    0.2

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    FULL BRIDGE MODEL  We link together the  n  cross sectionswhich are subject to the restoring force of the hangers.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    60/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    FULL BRIDGE MODEL  We link together the  n  cross sectionswhich are subject to the restoring force of the hangers.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    61/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    FULL BRIDGE MODEL  We link together the  n  cross sectionswhich are subject to the restoring force of the hangers.

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    62/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    FULL BRIDGE MODEL  We link together the  n  cross sectionswhich are subject to the restoring force of the hangers.

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    63/95

    We take  m =  = 1, we put  y 0  = y n+1  = θ0  = θn+1  = 0, and we

    obtain a system of 2n  second order equations:

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    FULL BRIDGE MODEL  We link together the  n  cross sectionswhich are subject to the restoring force of the hangers.

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    64/95

    We take  m =  = 1, we put  y 0  = y n+1  = θ0  = θn+1  = 0, and we

    obtain a system of 2n  second order equations:13 θ̈i  + U θi (Θ, Y ) = 0

    ÿ i  + U y i (Θ, Y ) = 0(i  = 1, ..., n) ,

    where (Θ, Y ) = (θ1 . . . , θn, y 1, . . . , y n) ∈ R2n and

    U (Θ, Y ) =n

    i =1

    F (y i  + sin θi ) + F (y i  − sin θi )

    +12

    ni =0

    K y (y i  − y i +1)2 + K θ(θi  − θi +1)

    2

    .

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We fix  n = 16 and  K y   = K θ  = 320. We choose initial data close toa longitudinal mode: the  k -th mode may be approximated by

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    65/95

    g y pp y

    ˙Y (0) = 0  ,   y i (0) = y  sin

    ik  π17

      (i  = 1, ..., 16)

    and  Θ̇(0) = Θ(0) = 0. The value of  y  measures the initialpotential energy (and hence the energy  E  of the system).These initial conditions give rise to a solution  Y   which   is not

    periodic; we seek a periodic solution “close” to  Y  and we denoteby T   = T (k , E ) its period.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We fix  n = 16 and  K y   = K θ  = 320. We choose initial data close toa longitudinal mode: the  k -th mode may be approximated by

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    66/95

    g y pp y

    ˙Y (0) = 0  ,   y i (0) = y  sin

    ik  π17

      (i  = 1, ..., 16)

    and  Θ̇(0) = Θ(0) = 0. The value of  y  measures the initialpotential energy (and hence the energy  E  of the system).These initial conditions give rise to a solution  Y   which   is not

    periodic; we seek a periodic solution “close” to  Y  and we denoteby T   = T (k , E ) its period.

    Then we perturb the torsion by taking initial data  θ̇i (0) and  θi (0)in the interval [−10−4, 10−4]. We define the transfer map

    Ψk E   :  R2n → R2n ,   Ψk E ( Θ̇(0), Θ(0)) = ( Θ̇(T ), Θ(T )) .

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We compute the eigenvalues  λi   = λi (k , E ) of  J Ψk E (0, 0), the

    Jacobian of the generalized Poincaré map at the origin

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    67/95

    Jacobian of the  generalized Poincare map  at the origin.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We compute the eigenvalues  λi   = λi (k , E ) of  J Ψk E (0, 0), the

    Jacobian of the generalized Poincaré map at the origin

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    68/95

    Jacobian of the  generalized Poincare map  at the origin.

    We display the plot of  E → max |λi |  for k  = 1, 2, 3, 4.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We compute the eigenvalues  λi   = λi (k , E ) of  J Ψk E (0, 0), the

    Jacobian of the generalized Poincaré map at the origin

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    69/95

    Jacobian of the  generalized Poincare map  at the origin.

    We display the plot of  E → max |λi |  for k  = 1, 2, 3, 4.

    In the damped version one sees similar behaviors.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    We compute the eigenvalues  λi   = λi (k , E ) of  J Ψk E (0, 0), the

    Jacobian of the generalized Poincaré map at the origin

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    70/95

    Jacobian of the  generalized Poincare map  at the origin.

    We display the plot of  E → max |λi |  for k  = 1, 2, 3, 4.

    In the damped version one sees similar behaviors.Is this a reliable method to determine the critical energy threshold?

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    71/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    72/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    73/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    74/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Define  critical energy threshold  E k  of the  k -th mode by

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    75/95

    E k  = inf E  > 0; max

    i |λi (k , E )| > 1

     .

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Define  critical energy threshold  E k  of the  k -th mode by

    | ( )|

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    76/95

    E k  = inf E  > 0; max

    i |λi (k , E )| > 1

     .

    The threshold  E k  depends on  k  and the  effective critical energythreshold  E  of the bridge satisfies

    E    ≤   min1≤k ≤n

    E k  .

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Define  critical energy threshold  E k  of the  k -th mode by

    E i f

    E 0 |λ (k E)| 1

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    77/95

    E k  = inf 

    E  > 0; max

    i |λi (k , E )| > 1

     .

    The threshold  E k  depends on  k  and the  effective critical energythreshold  E  of the bridge satisfies

    E    ≤   min1≤k ≤n

    E k  .

    Our explanation of the TNB collapse.  The TNB has collapsedbecause on November 7, 1940, the wind inserted enough energy to

    overcome the critical energy threshold. That amount of energycreated an internal resonance which was the starting spark for theappearance of destructive torsional oscillations.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    But this is a conference on  variational methods!

    http://goforward/http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    78/95

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Together with Alberto Ferrero (UniPMN) we set up a  continuousmodel for suspension bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    79/95

    model for suspension bridges.

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Together with Alberto Ferrero (UniPMN) we set up a  continuousmodel for suspension bridges

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    80/95

    model for suspension bridges.

    Prior works by McKenna, Walter, Lazer, Micheletti, Pistoia,Saccon...

    F. Gazzola - DipMat - PoliMi   Maths & Bridges

    Together with Alberto Ferrero (UniPMN) we set up a  continuousmodel for suspension bridges.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    81/95

    model for suspension bridges.

    Prior works by McKenna, Walter, Lazer, Micheletti, Pistoia,Saccon...

    We view the bridge as a plate

    Ω = (0, π) × (−, ) ( π)

    which is hinged on the small edges and free on the large edges:

    F. Gazzola - DipMat - PoliMi Maths & Bridges

    According to the  Kirchhoff-Love  theory of elasticity the bendingenergy of the plate subject to an external load   f    ∈ L2(Ω) is

    1 2 2

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    82/95

    E(u ) =  Ω1

    2(∆u )2 + (1 − σ)(u 2xy  − u xx u yy ) − fu   dxdy 

    where  σ  ∈ (0, 1/2) is the Poisson ratio. Due to the free part of theboundary, we neglect here the stretching energy.

    F. Gazzola - DipMat - PoliMi Maths & Bridges

    According to the  Kirchhoff-Love  theory of elasticity the bendingenergy of the plate subject to an external load   f    ∈ L2(Ω) is

    1 2 2

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    83/95

    E(u ) =  Ω1

    2(∆u )2 + (1 − σ)(u 2xy  − u xx u yy ) − fu   dxdy 

    where  σ  ∈ (0, 1/2) is the Poisson ratio. Due to the free part of theboundary, we neglect here the stretching energy.

    For  σ ∈ (−1, 1) the quadratic part of the functional  E(u ) is

    positive. We define the space

    H 2∗ (Ω) :=

    w  ∈ H 2(Ω);  w  = 0 on  {0, π} × (−, )

    which is a Hilbert space when endowed with the scalar product

    (u , v )H 2∗(Ω)  :=

     Ω

    [∆u ∆v  + (1 − σ)(2u xy v xy  − u xx v yy  − u yy v xx )]  dxdy  .

    F. Gazzola - DipMat - PoliMi Maths & Bridges

    By the Lax-Milgram Theorem, for any   f    ∈ L2(Ω) there exists aunique  u  ∈ H 2∗ (Ω) such that

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    84/95

     Ω

    [∆u ∆v  + (1 − σ)(2u xy v xy  − u xx v yy  − u yy v xx )] = Ω

    fv 

    for all  v  ∈ H 2∗ (Ω).

    F. Gazzola - DipMat - PoliMi Maths & Bridges

    By the Lax-Milgram Theorem, for any   f    ∈ L2(Ω) there exists aunique  u  ∈ H 2∗ (Ω) such that

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    85/95

     Ω

    [∆u ∆v  + (1 − σ)(2u xy v xy  − u xx v yy  − u yy v xx )] = Ω

    fv 

    for all  v  ∈ H 2∗ (Ω).

    The  strong form of this problem reads

    ∆2u  = f  

    u (0, y ) = u xx (0, y ) = u (π, y ) = u xx (π, y ) = 0

    u yy (x , ±) + σu xx (x , ±) = u yyy (x , ±) + (2 − σ)u xxy (x , ±) = 0.

    for, respectively, (x , y ) ∈ Ω,  y  ∈ (−, ),  x  ∈ (0, π).

    F. Gazzola - DipMat - PoliMi Maths & Bridges

    What can be done for this problem?

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    86/95

    F Gazzola - DipMat - PoliMi Maths & Bridges

    What can be done for this problem?•  If   f    = f   (x ) then one can separate variables and find explicitsolutions through Fourier series (Navier 1823, Lévy 1899, Zanaboni1940) This is possible only because two opposite edges are hinged

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    87/95

    1940). This is possible only because two opposite edges are hinged.The beam equation is somehow the limit of the plate equation as → 0.

    F Gazzola - DipMat - PoliMi Maths & Bridges

    What can be done for this problem?•  If   f    = f   (x ) then one can separate variables and find explicitsolutions through Fourier series (Navier 1823, Lévy 1899, Zanaboni1940). This is possible only because two opposite edges are hinged.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    88/95

    1940). This is possible only because two opposite edges are hinged.The beam equation is somehow the limit of the plate equation as → 0.

    •  Nonlinear restoring forces due to the hangers and the failure of the Hooke law, suggest to study the nonlinear problem

    ∆2

    u  + Υ(x )g (u ) = f   (x ) in Ω;

    here one should think of Υ to be the characteristic function of some proper subset  ω ⊂ Ω, the set where the hangers act. Theproblem is well-posed.

    F Gazzola - DipMat - PoliMi Maths & Bridges

    What can be done for this problem?•  If   f    = f   (x ) then one can separate variables and find explicitsolutions through Fourier series (Navier 1823, Lévy 1899, Zanaboni1940). This is possible only because two opposite edges are hinged.

    http://find/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    89/95

    ) p y pp g gThe beam equation is somehow the limit of the plate equation as → 0.

    •  Nonlinear restoring forces due to the hangers and the failure of the Hooke law, suggest to study the nonlinear problem

    ∆2

    u  + Υ(x )g (u ) = f   (x ) in Ω;

    here one should think of Υ to be the characteristic function of some proper subset  ω ⊂ Ω, the set where the hangers act. Theproblem is well-posed.

    •   Introducing the kinetic energy, yields the “hyperbolic” problem

    u tt  + ∆2u  + δ u t  + Υ(x )g (u ) = f   (x ) in Ω × (0, T );

    existence, uniqueness, regularity, asymptotic behavior...

    F Gazzola - DipMat - PoliMi Maths & Bridges

    What about the  eigenvalues and eigenfunctions  of the linearoperator?   ∆2u  = λu  and (bc).

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    90/95

    F Gazzola - DipMat - PoliMi Maths & Bridges

    What about the  eigenvalues and eigenfunctions  of the linearoperator?   ∆2u  = λu  and (bc).

    The spectrum is made by a sequence (in fact many sequences) of

    http://find/http://goback/

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    91/95

    The spectrum is made by a sequence (in fact, many sequences) of 

    eigenvalues

    (1 − σ)2 < λ1  < λ2  

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    92/95

    The spectrum is made by a sequence (in fact, many sequences) of 

    eigenvalues

    (1 − σ)2 < λ1  < λ2  

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    93/95

    The spectrum is made by a sequence (in fact, many sequences) of 

    eigenvalues

    (1 − σ)2 < λ1  < λ2  

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    94/95

    The spectrum is made by a sequence (in fact, many sequences) of 

    eigenvalues

    (1 − σ)2 < λ1  < λ2  

  • 8/16/2019 New Mathematical Models for Suspension Bridges - Gazzola (Alghero, 2013)

    95/95

    THANK YOU FOR YOUR ATTENTION!

    F Gazzola DipMat PoliMi Maths & Bridges

    http://find/