on the time-domain analysis of mechanical systemssbel.wisc.edu/documents/presentation06aug30.pdf ·...
TRANSCRIPT
![Page 1: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/1.jpg)
On the Time-Domain Analysis of Mechanical Systems
Dan NegrutDept. of Mechanical EngineeringUniversity of Wisconsin, Madison
![Page 2: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/2.jpg)
Talk Overview
� The need for time-domain analysis� Multibody Dynamics in the CAE landscape
� On the implicit integration of the index-3 DAE of Multibody Dynamics
� The HHT method
� A stabilized index-2 approach for Multibody Dynamics simulation
� Conclusions
![Page 3: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/3.jpg)
The Virtual Prototyping Problem
� Product Life Cycle� GOAL: reduce (ideally replace)
hardware prototyping with virtual prototyping
� Virtual Prototyping
� Cheaper
� Faster
� But… captures real world phenomena only to a certain degree (from where the need for some physical testing)
![Page 4: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/4.jpg)
Multibody Dynamics (MD) and CAE
� MD as end goal:
� J-turn at 65mph: will the vehicle roll over?
� What design parameters do I need to change to improve performance?
CAE
NVH Nastran, ANSYS
Durability&Fatigue Analysis ABAQUS, Marc, ANSYS
Crashworthiness LS-Dyna, Radios, Dytran
Multibody Dynamics ADAMS, LMS-Virtual Lab, SimPack
CAD CATIA, Pro/E, Solid/Edge
� MD as intermediate step:
� Compute load history for fatigue analysis
� Interface to control system (mechatronics)
![Page 5: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/5.jpg)
Example: Multibody Dynamics Analysis
� Track simulation for performance
improvement
![Page 6: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/6.jpg)
Multibody Dynamics:What Are We Interested In?
� Generalized positions :
� Generalized velocities:
� Generalized accelerations:
� Reaction Forces (Lagrange Multipliers):
� Action (Applied) Forces:
� User Defined Variables:
� Solution and its Time Derivative for user-defined First Order Initial Value Problems (coming from Controls)
![Page 7: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/7.jpg)
Differential Algebraic Equations of
Multibody Dynamics
![Page 8: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/8.jpg)
HHT IntegratorHilber – Hughes – Taylor (HHT), 1977
� Designed for time evolution of linear finite element problem
� The idea: advance the simulation (solution) in time to find evolution of
the mechanical system
![Page 9: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/9.jpg)
HHT Integrator(Contd.)
� The HHT Method:
� 2nd Order Accurate and A-Stable
� Attractive numerical damping properties controlled
through the α parameter
![Page 10: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/10.jpg)
Discretization of DAE Equations via HHT
� The Discretized Equations Solved at each Time-Step tn+1:
![Page 11: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/11.jpg)
Solving the Discretization Non-Linear System
� Quasi-Newton Approach to Find the Solution. Corrections Computed as :
� Correction is then applied as:
![Page 12: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/12.jpg)
Are we there yet?
� Questions that remain to be answered:
� Q1: How do you know whether the solution is accurate
enough?
� Q2: How do you choose the integration step-size h?
� Q3: How many Newton corrections do you take at each
time step before you stop?
![Page 13: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/13.jpg)
Notation
– dimensionless constant
– number of error controlled states
– acceptable integration error (user specified)
![Page 14: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/14.jpg)
Local Error Estimation
� Local Integration Error (Definition):
� Local Integration Error (Asymptotic Expansion Result):
→ Exact solution
→ Numeric solution
![Page 15: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/15.jpg)
� Equivalently: , where
Q1: The Accuracy Question
� Local Integration Error:
� Composite Integration Error defined as:
� Integration step successful if:
![Page 16: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/16.jpg)
Q2: The Step Size Selection Question
� I’d like be such that :
� This leads to
![Page 17: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/17.jpg)
� Question: How many iterations should I take?
� Iterative Process Stopping Criteria ( s2=0.1% ):
Q3: Non-Linear Solver Stopping Criteria
![Page 18: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/18.jpg)
� Quasi-Newton approach, with linear rate of convergence ξ. Therefore,
� The stopping criteria leads to
Q3: Non-Linear Solver Stopping Criteria (Contd.)
![Page 19: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/19.jpg)
Summary of Key Formulas
� Notation:
� Solution is accurate provided: (Q1: Error Control Question)
� Select the next integration step as:(Q2: Step-Size Selection Question)
� Take Newton iterations until:(Q3: Stopping Criteria Question)
![Page 20: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/20.jpg)
Putting Things in Perspective…
� Multibody Dynamics in the CAE landscape
� HHT-I3 integrator proposed to determine the
time evolution of a mechanical system� Implicit integrator: differential problem transformed into
non-linear algebraic problem
� Three integrator questions were answered
![Page 21: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/21.jpg)
Numerical Results
� Validation:
� Belt Model
� 150 Bodies
� Dimension of the Problem: 3923 equations
� Length of the Simulation: 200 ms (more than two revolutions of the belt)
![Page 22: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/22.jpg)
Numerical Results Comparison: (vs. BDF Stabilized Index 2)
� Tooth 57; X-translational velocity
![Page 23: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/23.jpg)
Numerical Results
� New Integrator:
� DAE Index 3 Approach
� Formula: HHT
� Variable step, constant order
� Integration Error: ε =1E-4
� Reference Integrator (ADAMS):
� Stabilized Index 2 DAE Approach
� Formula: Implicit BDF
� Variable step, variable order
� Integration Error: ε=1E-2
� CPU Time: 1386.7 seconds� CPU Time: 5898.3 seconds
![Page 24: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/24.jpg)
CPU Comparison
24h 45min79h
HHTGSTIFF
� Crawler Model� P4-2.4GHz, 512Mb, WinXP� Tend = 10 sec
![Page 25: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/25.jpg)
CPU Comparison
� Psa_3d_4000� 15689 equations� Tend = 75msec 1972.417656.1
HHTSI2 BDF
2537.610250.0
HHTSI2 BDF� A_4500
� 16258 equations� Tend = 120msec
![Page 26: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/26.jpg)
HHT on Wall Street…
Santa Ana, CA – MSC.Software Corp. (NYSE: MNS), the leading global provider of virtual
product development (VPD) products including simulation software and services, today
announced the release of MSC.Software SimOffice products MSC.Nastran 2005, MSC.Patran
2005, MSC.ADAMS 2005, MSC.Marc 2005 and MSC.Dytran 2005.
……
MSC.ADAMS 2005: New ADAMS/Engine Piston module for understanding secondary
piston motion; belt and chain solutions in ADAMS/Engine are 4 to 12 times faster; breakthrough productivity gains for complex durability and solid contact models using new 2nd order Hilber-Hughes-Taylor integrator; critical stresses can now be obtained at least a factor of 3 times faster…
![Page 27: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/27.jpg)
What the Wall Street release didn’t mention…
� There is yet no formal analytical proof for the
proposed integration method:
� Recall that HHT method originally proposed for second order linear ODEs
� For proposed index-3 DAE HHT-based approach:
� No global convergence analysis
� For some simulations frequent integration step-size changes led to instabilities
![Page 28: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/28.jpg)
Instability due to frequent step-size change
� Planetary gear model
� Frequent step-size changes leads to
� Order reduction
� Instabilities
� New method (HHT-SI2) proposed to address drawbacks associated with HHT-I3
![Page 29: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/29.jpg)
HHT-SI2: Stabilized index-2 HHT method
� Joint work with Laurent Jay, of the University of Iowa
� Method is more expensive, but addresses the fundamental issues associated with the HHT-I3
method
� Formal global convergence proof
� Method is of order 2 in position and velocity
� Algorithm can change step-size during simulation without
any instability (global convergence order is maintained)
![Page 30: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/30.jpg)
Proposed Approach: Preliminaries
� Improve accuracy of numerical solution by
considering stabilized index 2 formulation
� Notation:
�Order reduction equation
�Euler-Lagrange equation
�Position kinematic equation
�Action-force acceleration
�Reaction-force acceleration
�Velocity kinematic equation
![Page 31: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/31.jpg)
Proposed Approach: The Method
� Discretization Formula
� Notation used:
� a1, Λ0, and Λ1: solution of nonlinear system
![Page 32: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/32.jpg)
HHT-SI2: Main Theoretical Result
� Proposed method is second order globally convergent
� More precisely, if
� Then (yn, zn, an) satisfy
� Proof is long and technical
![Page 33: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/33.jpg)
•• mm11 = mass of the pendulum= mass of the pendulum
•• LL11 = half the length of the pendulum= half the length of the pendulum
•• kk11 = spring stiffness= spring stiffness
•• cc11 = damping coefficient= damping coefficient
•• g = gravityg = gravity
Numerical Experiments 1:Simple Pendulum
![Page 34: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/34.jpg)
Order Analysis: Simple Pendulum
Error of extended HHT: Pendulum α =- 0.3
-9.5
-8.5
-7.5
-6.5
-5.5
-4.5
-4.5 -4.3 -4.1 -3.9 -3.7 -3.5 -3.3 -3.1 -2.9
h
err
ors
in
an
gle
an
d a
ng
ula
r v
elo
cit
y
Error of extended HHT: Pendulum, α = 0
-10
-9
-8
-7
-6
-5
-4
-4.5 -4.3 -4.1 -3.9 -3.7 -3.5 -3.3 -3.1 -2.9
h
err
ors
in
an
gle
an
d a
ng
ula
r v
elo
cit
y
αα = = --0.30.3 αα = 0= 0
![Page 35: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/35.jpg)
Numerical Experiments 2:Double Pendulum
•• m1 = mass of the pendulumm1 = mass of the pendulum
•• m2 = mass of second pendulumm2 = mass of second pendulum
•• L1 = half the length of the pendulumL1 = half the length of the pendulum
•• L2 = half length of the second pendulumL2 = half length of the second pendulum
•• k1 = spring stiffness of first pendulumk1 = spring stiffness of first pendulum
•• k2 = spring stiffness of second pendulumk2 = spring stiffness of second pendulum
•• c1 = damping coefficient first pendulumc1 = damping coefficient first pendulum
•• c2 = damping coefficient second pendulumc2 = damping coefficient second pendulum
•• g = gravityg = gravity
![Page 36: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/36.jpg)
Order Analysis:Double Pendulum
![Page 37: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/37.jpg)
Numerical Experiments 3:Comparison with Analytical Solution
� Test Problem
� For considered initial conditions, analytical
solution is
![Page 38: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/38.jpg)
Order Behavior
![Page 39: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/39.jpg)
Variable Stepsize:Without modification, reduction to order 1
![Page 40: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/40.jpg)
Proposed modification for variable stepsize
![Page 41: On the Time-Domain Analysis of Mechanical Systemssbel.wisc.edu/documents/presentation06Aug30.pdf · On the Time-Domain Analysis of Mechanical Systems Dan Negrut Dept. of Mechanical](https://reader035.vdocuments.us/reader035/viewer/2022081517/5b2a16c67f8b9aa1058b7ea9/html5/thumbnails/41.jpg)
Putting things in perspective…
� Purpose of work:� Provide methods to determine time evolution of mechanical systems
� Two methods were proposed/discussed:
� HHT-I3� Adapted from structural dynamics (Hilber-Hughes-Taylor, 1977)
� Very fast
� Accuracy and stability an issue at times
� No formal order and global convergence proofs yet
� HHT-SI2� Includes additional equations (velocity constraint equations)
� Slower
� More robust
� Formal convergence proofs, supports variable step-size integration