Modeling Dynamic Effects

 Our work so far has been time-independent

• Structural problems – static (no acceleration)
• Thermal problems – steady state (no change with time)
 Structural dynamics
• We will now consider problems in which time plays a key role
• Various forces lead to structural acceleration
• Structural response varies with time
 Several ways of considering structural dynamic response
• Modal analysis – natural frequencies of a structure
• Harmonic analysis – response of structure under harmonic loading
• Transient analysis – response of structure under general loading
 Simplifications for finite element analysis
• First two have harmonic (sinusoidal) displacements in time
• Governing equations can be simplified significantly (ME 512)
• Third requires transient analysis methods (ME 612)
9-1

1
Importance of Understanding Resonance
 Resonance
• Structure vibrates excessively under periodic loading
• Most typically associated with structure’s natural frequency
• Several examples of importance of resonance

9-2

2
 Structural Resonance
 Tacoma Narrows Bridge – famous resonance example
• Wind flow transverse to the bridge
• Torsional oscillations leading to structural failure
• Failure believed related to flutter in aircraft not vortex shedding

http://www.engineering.wright.edu/~jslater/SDTCOutreachWebsite/aeroelasticity.htm 9-3

3
 Resonance – Tacoma Narrows Bridge
 Discussion from Wikipedia

9-4

4
 Aircraft Flutter
 Interaction between air flow and aircraft structure
• Leads to vibration of structure
• Can be catastrophic if increases with time
• Study using wind tunnels / advanced analysis methods

Boeing 747 C5 Tail

http://www.vibrationdata.com/wind.htm
9-5

5
 Vortex Shedding
 Cylinder in cross–flow
• Periodic shedding of vortices (von Karman)
• Leads to periodic loading of cylinder
• Important for tall, cylindrical objects such as smoke stacks / towers
– Avoid using helical strakes to disturb generation of vortices

http://www.idom.es/ada/References/ref-rain-radar/ref-rain-radar.htm#Von%20Karman%20Vortex%20Shedding
http://www.engineering.wright.edu/~jslater/SDTCOutreachWebsite/aeroelasticity.htm 9-6

6
 Engine Valvetrain Resonance
 Design of engine valvetrain
• Engine valves – springs push valves close, push rods open valves
• First inclination – stiffer springs mean better response
• Actual result – stiffer springs are heavier, worse response
• Note result at below right – harmonic analysis
– Vary frequency and determine resulting displacement

http://www.stockcarracing.com/techarticles/general/scrp_0608_valvetrain_myths/
9-7

7
 Spring-Mass-Damper System
 Consider a 1D spring-mass-damper system
• Spring stiffness k
• Damping coefficient c
• Mass m
• Displacement u(t)
• Force r(t)

 Draw free body diagram

of forces acting on the mass
• Sum of forces equals 0
• Include inertial forces (acceleration)

F 0 

9-8

8
 FEA For Structural Dynamics
 For 1D spring-mass-damper
system we found

 We write a similar expression for FEA

• Nodal DOFs (displacement) D and forces r
• Mass matrix M, damping matrix C, stiffness matrix K

 Note simplification for static case (ME 512 so far)

• Forces and displacements are no longer time-dependent

9-9

9
 Mass Matrix
 Fundament FEA equation of structural dynamics

 Two new matrices to consider

• Mass matrix M and damping matrix C
• ME 512 – assume no damping (C = 0)

 Mass matrix M  
• Consider D  D  0  M D  R
• Recall direct method for determining K matrix (K D = R)
– Set jth term in D equal to 1, all others to 0
– Forces R that resulted are jth column of K
• Same approach for

finding M
– Set jth term in D equal to 1, all others to 0
– Forces R that resulted are jth column of M
• How much force is required to achieve 1 unit acceleration at a node?
9-10

10
 Mass Matrix – 1D Beam Example
 Consider a 1D beam element
• Length L, Area A, Density 
• Leads to total mass AL
 Consider mass matrix terms

• 4 acceleration terms (D )
• 4 force terms (R)
• M is a 4  4 matrix (same size as K)
 m11 m12 m13 m14   v1   F1 
m m22 m23 m24  1   M 1 
  R   21
MD   
 m31 m32 m33 
m34 v2   F2 
 
m41 m42 m43 m44  2  M 2 
 As with stiffnesss matrix, can find mass matrix two ways
• Direct method and analytical method
• Several assumptions can be used to simplify mass matrix as desired
9-11

11
 1D Beam Example – Lumped Mass
 Lumped mass matrix
• Distribute all mass terms to the nodes
• Simplest approximation for 1D beam element
 One approach – ignore rotational inertia effects
• Beam modeled by a point mass at each end

 
MD

 Second approach – add rotational inertia effects

• Include rotational inertia of a solid beam from classical physics

 
MD

9-12

12
 Consistent Mass Matrix
 Lumped mass matrices have certain benefits
• Beneficial for some transient analyses because no off-axis terms
• Represents an approximation of the acceleration behavior
 Proper form is the consistent mass matrix
• Acceleration considered based on response of every point in body
• This is accomplished by the acceleration field at all points in body
– Based on the shape functions (N) and nodal accelerations ( D)
• Leads to an analytical equation for calculating mass matrix
– Similar to equation for K, result is a fully populated mass matrix

Remember to
make sure that
 156 22 L 54  13L   v1 
density, force
   A L
 22 L
 4 L2
13L  3L2  1  and distance
MD   units all agree!
420  54 13L 156  22 L  v2 
 
 13L  3L
2
 22 L 4 L2  2 
9-13

13
Free Vibration – 1D Spring–Mass Example
 1D spring-mass-damper example
• Set c = 0 (no damping) and no force
• Equation of motion becomes
k u  m u  0 (referred to as free vibration)

 Differential equation is easily solved

• An arbitrary phase angle () added to the solution is also a solution
k u  m u  0  u (t )  u0 sin  t
 Substitute back into original equation to find
k u0 sin  t   m  u0  2 sin  t   0  k   2 m u0 sin  t  0
 This equation has two solutions
• Trivial solution: u0 = 0 (k, m are arbitrary)
• Free vibration:   k m (u0 is arbitrary)
• Refer to  as the natural frequency (units are radians/second)
9-14

14
 Free Vibration – General FEA Equations
 Recall general FEA equation of motion
KD  C D M D   R
 Set C = 0 (undamped) and R = 0 (unforced)
KD  M D   0
 As with 1D case, this has a simple solution

Dt   D sin  t  K D sin  t   M   2 D sin  t  0 
K   
M D sin  t  0
2

 Again, this equation has two solutions

• Trivial solution: D = 0 (K, M are arbitrary)
• Free vibration:  is the natural frequency
 The latter represents an eigenvalue problem
• Matrix equation – cannot solve algebraically like spring-mass case
• Eigenvalue (2), eigenvector D
9-15

15
 Eigenvalue, Eigenvector – Mathematics
 A bit of background on eigenvalue / eigenvector problems
• Information only – not needed for final exam!
 Consider a matrix A that multiplies a vector x
• This leads to another vector b (i.e. A x = b)
 Are there any cases when b is a multiple of x (b =  x)?
• Substitute and rearrange the equation to find
• Solutions: 1) x = 0 (trivial) 2)   0 (eigenvalue), x  0 (eigenvector)
A x   x  A   I x  0
 How does this relate to our case? Premultiply by M-1 or K-1
• Clearly M–1 K is A, 2 is  (eigenvalue) and D is x (eigenvector)

K    
M D  0  M 1 K   2 I D  0
2

 How do we find eigenvalues, eigenvectors (either form is fine)
• Eigenvalue when (…) term is singular  det (…) = 0
• Once eigenvalue known, eigenvector makes equation become 0
• Kreyszig handout (book used in ME 565, ME 566 – Adv Engr Math) 9-16

16
 Free Vibration and FEA – Several Points
Eigenvalue Problem:  K   2
M D  0
 The magnitude of D is irrelevant
• Any multiple of D is also an eigenvector (i.e. solves problem)
• Scale is often set using (benefits for other analyses)
 Do not need to prevent rigid body motion cases in ANSYS
• First several modes will represent rigid body motion modes if they exist
• Associated frequencies will be (approximately) 0
• Note that you must include any actual supports to get correct results!
 Often interested in lowest non-zero frequency
• Called the fundamental natural frequency of the system
• Typically design system to operate at frequencies below this value
 Note that natural frequency can be expressed two ways
• As , with units of radians/second
• As circular frequency f, with units of Hertz ( = 2 f)
• ANSYS returns frequencies as f in Hertz (Hz) 9-17

17
 Beam Resonance Example
 Solve for natural frequencies of simple beam problem
• 6 DOFs total, restrain 4 or 5 of them
• Determine natural frequencies and associated mode shapes

9-18

18
 Natural Frequency, Damping
 Consider structure operating under cyclic load (L)
• FEA finds fundamental natural frequency 1
• When is static analysis acceptable?
• When do you need to perform a dynamic analysis?
 Good approximation: assume static if L < 1 / 3
• Results are generally within 10% of dynamic analysis
– Perform a harmonic analysis to include dynamic effects (next topic)
• As you approach 1, forces of inertia become important
• Structure operated near natural frequency can have very short life
– Large displacements, large stresses  can lead to rapid failure
 Effect of damping
• Most structures have relatively little damping
• Assume undamped free vibration in most cases
– If included, will find 1-Damped < 1-Undamped (but only slightly different)
• ANSYS can perform calculations using damping
– Based on complex analysis methods [i.e. v = p + i q, i = (-1)½]
– Only do this if you are sure you know what you are doing! 9-19

19
 Force, Distance, Mass Units
 As discussed previously, make sure units are consistent
• See NASA handout from earlier in semester
• Before performing dynamic analysis, always check your units!
• Two common metric unit approaches below

• English units – work through similar exercise to check

– Usually simply divide density by appropriate value of g
9-20

20
 Structural Symmetry
 Many structures under consideration have symmetry
• Stress analysis – use symmetry to simplify problem
 Vibration modes are not necessarily symmetric
• Example – vibrating string that is simply supported
• Structure and loading (none) are symmetric about midplane
• 1st vibration mode is symmetric but 2nd vibration mode is not

 In general, do not use symmetry in free vibration analyses

– Will result in loss of non-symmetric vibration modes
 Demonstration – consider example two ways
• 24” square steel plate, 0.10” thick
• All edges simply supported
• Solve using ¼ symmetry
• Repeat using full model
9-21

21
 Example Problem – ¼ Symmetric Model
 Boundary conditions for ¼ symmetric model

9-22

22
 Example Problem – ¼ Symmetric Model
 1st non-zero frequency (33.51 Hz) for ¼ symmetric model

9-23

23
 Example Problem – ¼ Symmetric Model
 2nd non-zero frequency (167.72 Hz) for ¼ symmetric model

9-24

24
 Example Problem – Full Model
 Boundary conditions for full model

9-25

25
 Example Problem – Full Model
 1st non-zero frequency (33.51 Hz) for full model

9-26

26
 Example Problem – Full Model
 2nd non-zero frequency (83.83 Hz) for full model

9-27

27
 Example Problem – Full Model
 5th non-zero frequency (167.72 Hz) for full model

9-28

28
 Nodal Reduction
 Eigenvalue problems can be computationally intensive
• Large models – may need way to improve solution speed
 Recall earlier discussion about static condensation
• Used for Q6 element to reduce internal DOFs

 Similar procedure can be used for modal analysis

• Only keep certain DOFs in your solution
• DOFs to keep – master DOFs | DOFs to condense – slave DOFs
• Can result in significant reduction of solution time
9-29

29
 Harmonic Response Analysis
 Undamped free vibration – previous slides
• No damping
• No forces applied – R(t) = 0
Eigenvalue Problem : K   2

M D sin  t  0
• Resulting solution: N modes at i with Di(t) = D i sin t

 Harmonic analysis
• No damping
• Force applied is harmonic – R(t) = R sin t
– Every nodal load is in phase
– Frequency can be any value
• Substitute into governing FEA equation
  R
K DM D
K   2

M D sin  t  R sin  t
K* D  R ; K*  K   2 M  
9-30

30
 Harmonic Response Analysis
Harmonic Analysis : K   M  D sin  t  R sin  t
2

K D  R ; K  K   M 
* * 2

 Harmonic response analysis

• Can consider effect of cyclic loading on structure
– Does not include transient part of solution (when load first applied)
• Accounts for inertial effects ignored in a static analysis
• No more complicated than static analysis in terms of solution
– Still looks like static K D = R fundament equation of FEA

 What happens when  = i (one of the natural frequencies)?

• Left hand side of equation is defined to be 0
• Right hand side of equation is defined to be nonzero
• There is no solution – cannot force structure in this way
– Displacements always increase with each cycle
– Structure will likely fail due to large displacements
– Need nonlinear time-dependent analysis to consider fully
9-31

31
 Harmonic Response Analysis
 Harmonic response analysis
• Usually performed at a series of frequencies
• Plot results at a given frequency (similar to static analysis plots)
• Can also plot various DOFs / nodal loads versus frequency
• Frame analysis problem from textbook

9-32

32
Transient Analysis – General Dynamic Loading
 Cases considered so far
• Modal analysis – undamped, unloaded
• Harmonic analysis – undamped, sinusoidally loaded
– Does not include transient portion of solution
 Transient analysis (not on final exam)
 M D
K DC D   R
• Loaded by general function R(t)
• Can be damped or undamped
• Generally two approaches to solution
 Modal superposition
• Solve for first M non-zero free vibration modes (M – your decision)
• Combine appropriately to predict solution
 Direct integration – full transient analysis (ME 612)
• Treat FEA equation as an initial value problem (diff. eq.)
• Solve at a series of points in time to predict response
9-33

33
 FEA Transient Dynamic Analysis
 Example – transient structural analysis slide from ME 612
• Solve problem as a series of time steps (t = t0, t0 + t, t0 + 2 t, etc.)
• Numerical integration approach – converts terms to mass / damping force vectors
• Shown for information only – do not need for ME 512 final exam!
  K D  R
MD
Displacement  u n  d(tn )
d
Velocity  u n  d(tn )
dt

~
K D  R  RM  RC

34
 Transient Analysis – Example
 Frame analysis problem from textbook
• Subject to impulse load (force for period of time)
• ANSYS – requires two different “load steps”
 Book solves problem both ways
• Modal superposition
• Direct integration

9-35

35
 Transient Analysis – Example
 Example
• Result at
at single node

 Modal analysis
• Number of
modes used
varies

 Direct integration
• Time step
used varies

 Note that sol’n

is fairly noisy
9-36

36
 ME 612 – FEA II
 Provide a brief overview of ME 612
• Topics / class style – below is Spring 2015 schedule

9-37

37