You are on page 1of 39

SPAÇAR: A Finite Element Approach in

Flexible Multibody Dynamics

UIC Seminar
Arend L. Schwab
Google: Arend Schwab [I’m Feeling Lucky]

September 27, 2004 University of Illinois at Chicago

Laboratory for Engineering Mechanics


Faculty of
Vermelding Mechanical
onderdeel Engineering
organisatie
Acknowledgement

TUdelft: UTwente:
Hans Besseling Ben Jonker
Klaas Van der Werff Ronald Aarts
Helmut Rankers … MSc students
Ton Klein Breteler
Jaap Meijaard
… MSc students

September 27, 2004 2


Contents

• Roots
• Modelling
• Some Finite Elements
• Eqn’s of Motion
• Examples
• Discussion

September 27, 2004 3


Engineering Mechanics at Delft

From Analytical Mechanics in 50’s:


Warner T. Koiter
On the Stability of Elastic Equilibrium, 1945
To Numercial Methods in Applied Mechanics in 70’s:
Hans Besseling
The complete analogy between the matrix equations and the continuous
field equations of structural analysis, 1963

September 27, 2004 4


Mechanism and Machine Theory

Application of Numerical
Methods to:

• Kinematic Analysis
• Type and Dimension
Synthesis
• Dynamic Analysis

CADOM project: Computer Aided Design of Mechanisms, 1972


Rankers, Van der Werff, Klein Breteler, Schwab, et al.

September 27, 2004 5


Mechanism and Machine Theory,
Kinematics
Denavit & Hartenberg, 1955

• Rigid Bodies
• Relative Coordinates (few)
• Kinematic Constraints (few)

September 27, 2004 6


Mechanism and Machine Theory,
Kinematics
Klaas Van der Werff, 1975

Finite Element Approach

• Flexible Bodies
• Absolute Coordinates and Large Rotations
(many)
• Kinematic Constraints = Rigidity of Bodies
(many)

Note: Decoupling of the positional nodes and the orientational nodes.

September 27, 2004 7


Multibody System Dynamics
Finite Element Approach
Key Idea:
Specification of Independent Deformation Modes of the Finite Elements

Coordinates:
(xp, p , xq , q) total 6

Deformation Modes:
total 6-3=3

1  x q  x p  l0
2   p  
yq  y p
tan  
xq  x p
3    q
September 27, 2004 8
Multibody System Dynamics
Finite Element Approach
Pro’s:
• Easy FEM assembly of the system equations
• Easy mix of partly Rigid and/or Flexible elements
• Small set of elements for Large class of Multibody Systems
• Absolute Coordinates and Large Rotations
• Gen. Deformation can act as Relative Coordinates
Con’s:
• Many coordinates, many constraints
• Non-Constant Mass Matrix

September 27, 2004 9


Multibody System Dynamics
Finite Element Approach
Generalized Deformation can act as Relative
Coordinates
Ex. Hydraulic Cylinder

1  flexible or input
 2  rigid
 3  rigid

September 27, 2004 10


Multibody System Dynamics

Compare to Rigid Bodies with Constraints


Milton Chace & Nicky Orlandea, DRAM, ADAMS, 1970

Rigid Bodies with Constraints FEM approach


Constraints are at the Joints Constraints are in the Bodies

September 27, 2004 11


3D Beam Element Coordinates:
(xp, p , xq , q) total 14

Deformation Modes:
total 14 – 6 = 8 – 2 = 6

1  l  l0
 2  l0 e zp  e qy  e yp  e qz / 2
l  xq - x p
p p  3   l  e zp ,  4  l  e qz
e  R ( λ )e y 0 
y
 5  l  e yp ,  6  l  e qy
 7  ( λ p  λ p  1) / 2 = 0
 8  ( λ q  λ q  1) / 2 = 0

Cartesian Coordinates xp = (x, y, z)p and Euler Parameters p =(0, 1, 2, 3)p
September 27, 2004 12
3D Hinge Element
Coordinates:
(p, q) total 8

Deformation Modes:
total 8 – 3 = 5 – 2 = 3

1  arctan 2e qz  e yp , e qz  e zp 
 2  e xp  e qz
 3  e xp  e qy
 4  ( λ p  λ p  1) / 2 = 0
 5  ( λ q  λ q  1) / 2 = 0
e yp  R ( λ p )e y0 

September 27, 2004 13


3D Truss Element

Coordinates:
(xp, xq ) total 6

Deformation Modes:
total 6 – 5 = 1

1  l  l0

September 27, 2004 14


3D Wheel Element
Coordinates:
(xw, p, xc ) total 10

Some Counting:
Pure rolling Rigid body has
3 degrees of freedom (velocities).

We need 10-3=7 Constraints on the Velocities.

Pure rolling is 2 Velocity Constraints, Lateral and


Longitudinal.

Leaves 7-2=5 Deformation Modes

September 27, 2004 15


3D Wheel Element
Coordinates:
(xw, p, xc ) total 10

Deformation Modes:
1  r  r0
2  ew  r
 3  g (x c )
 4  (r  e w )  n
 5  ( λ p  λ p  1) / 2
Generalized Slips:
s1  a  v c
v c  x w  ω  r
s2  b  x c
September 27, 2004 16
Ex. Universal or Cardan Joint

Physical Model

Two FEModels: (a) with 4 Rigid Hinges, and (b) with 2 Flexible Hinges

September 27, 2004 17


Ex. Universal or Cardan Joint

Two FEModels: (a) with 4 Rigid Hinges, and (b) with 2 Flexible Hinges

September 27, 2004 18


Dynamic Analysis
In the spirit of d’Alembert and Lagrange we transform the DAE

 M ij Dc ,i   x j   f i a  f i e 
D     
 c, j 0cc   c   Dc , jk x j x k 
in terms of generalized independent coordinates qj
with xi=Fi(qj) resulting in

Fi ,k M ij F j ,l ql  Qi  Fi ,k ( f i a  f i e  M ij F j ,kl q k q l )
From which we solve
ql
and Numerically Integrate as an ODE.

Note: the Elastic Forces are according to and

f i e  De,i e  e  Sef D f
September 27, 2004 19
Ex. ILTIS Road Vehicle Benchmark
- Rigid Cabin
- 4 Independently
Suspended Wheels
- CALSAP Tire Model

The ILTIS Vehicle

85 Elements
239 Gen. Def.
70 Nodes
226 Gen. Coord.
10 DOF’s
Suspension FEM Model
September 27, 2004 20
Ex. ILTIS Road Vehicle Benchmark

Static Equilibrium Results

September 27, 2004 21


Ex. ILTIS Road Vehicle Benchmark

Handling Performance Test:


Ramp-to-Step Steer Manoeuvre at v = 30 m/s.

CALSPAN tire model

Zero Lateral Slip

September 27, 2004 22


Ex. Slider-Crank Mechanism

Slider-Crank Mechanism from Song & Haug, 1980

Rigid Crank,Flexible Connecting Rod


=150 rad/s, 2% damping

Transient Solution

Periodic Solution

First Eigenfrequency of pinned joint


connecting rod 0= 832 rad/s

September 27, 2004 23


Linearized Equations of Motion
Equations of Motion can be Analytically Linearized at a Nominal Motion

 d  C q d  Kq d  f d
Mq
Even for Systems having Non-Holonomic Constraints!

 d  C q d  K d q d  K k q k  f d
M q
q k  Aq d  B d q d  B k q k
with: M: reduced Mass Matrix
C: Tangent Velocity dependant Matrix
K: Tangent Stiffness Matrix
qk: Kinematic Coordinates variations
A: Non-Holonomic Constraints
B: Tangent Reonomic Constraints Matrix

September 27, 2004 24


Ex. Slider-Crank Mechanism
Nominal Periodic Motion and small Vibrations described by the Linearized
Equations of Motion

Slider-Crank Mechanism from Song & Haug, 1980

Rigid Crank,Flexible Connecting Rod


=150 rad/s, 2% damping

Transient Solution

Periodic Solution

September 27, 2004 25


Linearized Equations of Motion at
Nominal Periodic Motion
Periodic Solutions for small Vibrations superimposed on a Nominal Periodic
Motion

Linearized Equations of Motion at Nominal Motion:

 d  C (t ) q d  K (t ) q d  f d (t )
M ( t ) q
The Coefficients in the Matrices are Periodic with Period T=2/
Transform these Matrices into Fourier Series:

~ ikt ~ ikt ~ ikt ~ ikt


(  M k e ) q
  (  Ck e )q  (  K k e ) q    fk e
d d d

k k k k
and assume a periodic solution of the form:
m
~ d eilt
q d   l
 q
l  m
September 27, 2004 26
Linearized Equations of Motion at
Nominal Periodic Motion
Periodic Solutions for small Vibrations superimposed on a Nominal Periodic
Motion

Substitution into the Linearized Equations of motion and balance of every individual
Harmonic leads to:

~ ~ ~ d  ~f
 k l
{K
l
 ( il ) C k l  ( l ) 2
M k l } q l k

These are (2k+1)*dof linear equations from which we can solve the
2k+1 harmonics: q ~d
l
Which form the solution of the small Vibration problem:
m
~ d eilt
q  d
 l
 q
l  m

September 27, 2004 27


Ex. Slider-Crank Mechanism
Slider-Crank Mechanism from Song & Haug, 1980

FEModel: 2 Beam Elements for


the Flexible Connecting Rod

Rigid Crank,Flexible Connecting Rod


=150 rad/s, 2% damping
Transient Solution
Periodic Solution

September 27, 2004 28


Ex. Slider-Crank Mechanism
Maximal Midpoint Deflection/l for a range of ’s

Damping 1% and 2%
First Eigenfrequency of pinned joint
connecting rod 0= 832 rad/s
Resonace at 1/5, 1/4, and 1/3 of 0

Linearized Results

Full Non-Linear Results

September 27, 2004 29


Ex. Slider-Crank Mechanism
Individual Harmonics of the Midpoint Deflection/l for a range of ’s

First Eigenfrequency
of pinned joint
connecting rod
0= 832 rad/s
Resonace at 1/5, 1/4,
and 1/3 of 0

Quasi Static Solution

September 27, 2004 30


Ex. Dynamics of an Uncontrolled Bicycle

Cornell University, Ithaca, NY, 1987: Yellow Bike in the Car Park

September 27, 2004 31


Ex. Dynamics of an Uncontrolled Bicycle

Cornell University, Ithaca, NY, 1987: Yellow Bike in the Car Park

September 27, 2004 32


Ex. Dynamics of an Uncontrolled Bicycle

Modelling Assumptions:
- rigid bodies
- fixed rigid rider
- hands-free
- symmetric about vertical plane
- point contact, no side slip
- flat level road
- no friction or propulsion

Note: This model is energy conservative

September 27, 2004 33


Ex. Dynamics of an Uncontrolled Bicycle
FEModel: 2 Wheels, 2 Beams, 6 Hinges

3 Degrees of Freedom:

 lean angle  
q d   steer angle  
 
rear wheel rot.  r 
4 Kinematic Coordinates:

 front wheel rot.  f 


 yaw angle rear frame 
q 
k 
 rear contact pnt. xr 
 
 rear contact pnt. y r 

September 27, 2004 34


Ex. Dynamics of an Uncontrolled Bicycle
Forward Full Non-Linear Dynamic Analysis with an initial small side-kick

Forward Speed:
v = 3.5 m/s
v = 4.5 m/s

September 27, 2004 35


Ex. Dynamics of an Uncontrolled Bicycle
Investigate the Stability of the Steady Forward Upright Motion by
means of the Linearized Equations of Motion at this Steady Motion

Linearized Equations of Motion for Systems having Non-Holonomic Constraints in State-Space


form:

 M 0 0   q  d   C Kd K k  q d  0
   d    d   
 0 I 0   q     I 0 0   q   0
 
 0 0 I   q k   A  B d  B   q  0
k k

Assume an exponential motion for the small variations:

 q  q e  t
September 27, 2004 36
Ex. Dynamics of an Uncontrolled Bicycle
Rootloci of  from the Linearized Equations of Motion with as a Parameter the
Forward Speed v

Asymptotically Stable in the


Speed Range:
4.1 < v < 5.7 m/s

September 27, 2004 37


Ex. Dynamics of an Uncontrolled Bicycle
What happens for v>5.7 m/s?

Forward Speed:
v = 6.3 m/s

September 27, 2004 38


Conclusions

- SPAÇAR is a versatile FEM based Dynamic Modeling System for Flexible


and/or Rigid Multibody Systems .
- The System is capable of modeling idealized Rolling Contact (Non-
Holonomic Constraints).
- The System uses a set of minimal independent state variables, which
avoid the use of differential-algebraic equations.
- The Equations of Motion can be Linearized Analytically at any given
state.

September 27, 2004 39

You might also like