Diapositiva 1

1
Theoretical Background 1/1 Grafe, 1998
Theoretical Background 1/3
Theoretical Background 2/3
Theoretical Background 1/3
Heath monitoring with OO approach

• Damage identification by iteratively solving a
  non-linear optimization procedure via least
  square algorithm
• F.E. model represents the reference structure
• A sensitivity matrix, and error vector built
  from the correlation of the FRF analytical and
  experimental
• Localized and small damage entity required
• Refined numerical model and high number of design
  parameters are needed
• by-step enhancement is proposed for reduction
  of both computational time and computer memory
  amount
• Numerical and experimental validation reported

Damage identification techniques based on the
evaluation of the change of an Output signal wrt
the reference Response Model
Damage identification techniques based on the
evaluation of the change of an Output
signal wrt the reference Modal Model, Response
Model, Sound, Ultrasound
Correlation functions

• Advantage of Output-Only technique for the
  estimate of the modal parameters
• Only the output time responses of the structure
  are employed
• Measurements of the input loads of the structure
  are not necessary
• The used output data are those of the structure
  in operative conditions ? save costs and time
• The approach is particularly convenient whenever
  the input is unknown
• Aerospace field aeroelastic phenomena
• Civil field vibrations of builds and bridges

X ,A Experimental, Analytical

• Various type of damage
• identifiable from
• the natural frequency shift
• pattern modification of mode shapes

Sensitivity of the system to changes in the
design parameters
Uncertainties minimized through reduction of
data handling and manipulation
Component disembark required

• Health of the structure monitored evaluating
  changing in design parameter related to mass and
  stiffness distribution

• Low sensitivity to damage level
• Fine tuning of the F.E. model
• Low accuracy of experimental data from
  estimating process

Theoretical Background 2/3
Theoretical Background 3/3
Heath monitoring with OO approach
Results experimental investigation 1/6
Results experimental investigation 2/6

• Localized and small entities of structural
  damage requires high number of
• DOFs in FEM
• Design parameters

Experimental Analysis based on Output Only
Differences between the reference and the actual
strucure

• Experimental analysis
• Modal impact
• Free-Free B.C.
• Freq. Band 0-640 Hz
• 4096 Spectral lines
• 81 DOFs (trasversal)

Dynamic Response model
Added masses

• 64 64 design parameters
• 12 Experimental Dofs considered

• Damage identification process divided into
  consecutive steps
• For each iteration step, only the most sensitive
  design parameters to actual dynamic difference
  are retained
• Those parameters could differ from one iteration
  to another
• Small structural changes identified with
  acceptable computational costs

Undamaged Structure
Damaged Structure

• Reduction of 20 of thickness at the center of
  the plate (corresponding to the 28 element)

Variation of global parameters

• Introducing the weighting matrices

• 0.6 average change in fn
• No effects on damping ratios

• minimizing the functional

• The method do NOT identify the correct parameter
  ( 28)

• The solution is given by

Damage Identification

• and therefore

Results experimental investigation 6/6
Experimental investigation OO test on the
undamaged structure
Results experimental investigation 4/6
Results experimental investigation 5/6

• Localization process identified correct damaged
  region for mass distribution, not for stiffness

• 64 design parameters
• 12 Experimental Dofs considered
• Procedure speed up using a sensitivity matrix
  built on a frequency band of 0,300 Hz (final
  dimensions 2450 64)
• After 4 steps, the number of design parameters
  useful to describe the damage condition reduces
  to 2

• Actual damage involves mostly the stiffness
  characteristics, the mass changes are H.O.T.

Convergency history of stiffness related design
parameter- Step 4
Effects on global parameters
Mass-related design parameters reduces the
stability of the numerical algorithm (Least
Square solution)

• Two adjacent elements identified ( 28 and 36)
  corresponding to the actual damage location
• reduction in the eigenfrequency shifts
• increase in the correlation among the FRFs

Experimental investigation damaged
structure
Experimental investigation comparison
Concluding Remarks

• The procedure has been developed and successfully
  applied to an aluminum plate
• Sensitivity-based approach (from structural
  updating discipline) enhanced with a by-step
  algorithm
• Reduction of numerical instability, from noisy
  data, computer memory, and computational time
  achieved
• Blind search is dangerous
• Design parameters MUST describe the topology of
  the actual damage
• Analysis speed up with a suitable frequency band

FREQUENCY SHIFT DAMAGED V.S. UNDAMAGED MODELS
0.88
Variation of thickness 5
FRF COMPARISON ON THE FIRST MEASUREMENT POINTS