Content uploaded by Guillaume Martin
Author content
All content in this area was uploaded by Guillaume Martin on Sep 29, 2015
Content may be subject to copyright.
Test/Assembly correlation (Plate + Cable Guide) Contact parameters updating
Component parameters updating (Plate)
Review of model updating processes used for brake components
G. Martin1,3, E. Balmès1,3, G. Vermot-des-Roches1, T. Chancelier2
1SDTools, 2Chassis Brakes International, 3Arts et Metiers ParisTech/PIMM
Updating protocol
Shape correlation
•MAC
•Quality (MACErr)
•Frequency Err
Measurements
•Laser
•Accelerometers
Component 1 model
Updated Component 2 model
Identification
•IDRC
•IDRM
•Quality
Topo corr
Material parameters
•Density
•Young modulus
Material
parameter
updating
Contact properties
•Surface
•Pressure
•Stiffness
Contact
Properties
updating
Shape correlation
•MAC
•Quality (MACErr)
•Frequency Err
Assembly
measurements
•Laser
•Accelerometers
Identification
•IDRC
•IDRM
•Quality
Topo corr
Assembly model
Geometry updating
•Reconstruction
•or Morphing
Test geometry
•COMET
FEM Geometry
•Catia
•Reconstruction
Component 1 geometry
Topo corr
Test Case
Drum brake plate with fixed-point
and cable guide
•Step 1 : Geometry updating (Plate)
•Step 2 : Component parameters
updating (Plate)
•Step 3 : Contact parameters
updating (cable guide – fixed point)
Geometry updating (Plate)
Nominal FEM
180.000 Nodes
Point Cloud (STL)
240.000 Points
Error MAP STL/Nominal
Error MAP STL/Reconstructed
Reconstructed FEM
Correlation :
• Assuming modes
come ordered
• Density updated from
FEM volume and
physical part weight
• Relative error in
frequency (no shapes) :
𝐽 𝐸 = 𝑚𝑒𝑎𝑛 ∆𝑓
𝑖
𝑓
𝑖
(𝐸)
•Updated geometry leads to
a lower error than Nominal
geometry
•A bias in model
parameter occurs if we do
not update geometry first
Topology Correlation
ICP algorithm
Observability matrix building
No
optimization
mean
error = 1.14 mm
std
error = 0.88 mm
max error =
5.34 mm
Optimization
mean
error = 0.82 mm
std
error = 0.58 mm
max error = 3.01 mm
Contact Property updating
•Up to 2mm of distance between
Nominal Geometry and measured point
cloud
•The volume of the reconstructed
geometry if 11% lower than the volume
of the Nominal model
Assembled Updated
Components Laser Measurement
300 Points
Collocated transfer identification
(accelerometers)
Shape Correlation
Contact classically defined as a
node to surface constraint
•Mode crossing
•One test modeshape fully
uncorrelated
•One FEM modeshape fully
uncorrelated
•Many correlated below 70%
•Using surfaces leads to threshold effects and
several full model resolution : long time
•Using reduced model with stiffness on the
surface is fast, fairly equivalent and smoother
Surface sets definition
Updating contacts using
Surfaces or Stiffness
Conclusion
•Efficient & systematic updating process
•Geometry & contact : keys to good results
Identification of transfers Shape Correlation
Stiffness chosen for the first
cable guide bending mode to
be close in frequency with the
experimental modal analysis
•No Mode crossing
•MAC matrix fully diagonal
•All values above 83%