Optimal integrated design procedure for ASAC

1
Optimal integrated design procedure for ASAC
Leopoldo de Oliveria, Maíra da Silva, Gregory
Pinte, Wim Desmet, P.Sas K.U.Leuven, division PMA
- Leuven, Belgium Peter Mas, Herman Van der
Auweraer LMS International - Leuven, Belgium
2
Scenario
Always increasing demands for Noise Reduction and
Sound Quality improvement Promising results
from research on Active Control and Intelligent
Materials
There has to be a systematic way of including
Integrated Active Solutions during Product
Development Eventually, this procedure should be
suitable for concurrent optimization
3
Objectives

• Develop a methodology to simulate a fully
  coupled vibro-acoustic system with sensors and
  actuators.
• Stick as much as possible to standard simulation
  tools (FEM, BEM, Simulink)
• The method should allow further reduction and
  simulation of fully coupled closed loop systems
• Foresee the possibility to run Optimization
• Realise an Integrated Optimization of Structure
  and Controller Concurrently
• automate communication between different purpose
  software (Structural / Acoustic / Control)

4
Objectives (cont.)

• Why fully coupled ?

In some practical cases, where the structural
vibrations are not significantly affected by the
presence of the fluid, the vibro-acoustic problem
can be treated as an uncoupled problem. In this
way the structural vibrations are just input to
the acoustic system.
However, the present objective is to minimize
noise transmitted between two cavities through a
flexible structure. Thus, the structure needs not
only to excite a cavity, but also to be
acoustically excited.
5
Study Case LMS Sound-brick

• Simplified car cavity in concrete
• Flexible firewall between Passenger and Engine
  Cavities

6
Study Case LMS Sound-brick

• The LMS Sound-brick setup allows a relatively
  simple modelling, without loosing the generality
• The flexible Firewall between the Engine and
  Passenger Cavity is the only structural element
  affecting the system dynamics. It is also the
  only system subject to control actuation.
• It suits perfectly as an example of how the
  active control system design could be realized
  concurrently with the structure design, in order
  to get the better out of both.

7
Modelling Procedure Overview
8
Modelling Procedure Adding Sensors/Actuators

• Sensors as accelerometers can be added as
  concentrated mass
• An inertia shaker can be modelled as
  concentrated mass connected by a spring/damper
  element
• If those effects are neglectable, they can be
  represented as idealised input forces and output
  accelerations (velocities or displacements)

9
Modelling Procedure Vibro-Acoustics
LMS.Sysnoise procedure
The computation of a coupled FE/FE model with
Sysnoise furnishes a coupled modal base for both
models, which allows the calculation of forced
vibro-accoustic response.
10
Modelling Procedure Structural FEM

• The flexible structure consists of the steel
  firewall
• 920 x 545 x 1.5 mm / Clamped edges

FEM (quad4) nodes 231 elem 200
11
Modelling Procedure Acoustic FEM

• The acoustic mesh consists of the two empty
  cavities
• PC 3400 x 1560 x 1270 mm / EC 900 x 1100 x
  750 mm

FEM (hexa8) nodes 29 763 elem 26
098 MaxFreq (6) 100 lt 514 Hz 20 lt
1069 Hz
12
Modelling Procedure State Space
Import to MatLab the coupled Modal Base include
the electrical DoFs and convert to State Space
13
Modelling Procedure State Space
State Space model allows time domains simulation
for open and close-loop
14
Control Strategy

• Collocated Velocity Feedback

We want to minimize the resultant pressure at the
passenger heads, by acting and sensing only on
the structure Stability may become an
issue Moreover, the main objective is do
demonstrate a general methodology, in principle
since the modes is in Simulink, the user is free
to choose the control strategy
Disturbance
Objective
Controller
15
Optimization Procedure Overview
Concurrently Optimization (structure and
controller simultaneously)
General Scheme
LMS OPTIMUS
16
Optimization Procedure Cost Function
Concurrently Optimization (structure and
controller simultaneously)
This integrated approach should lead to the
global optimum solution, since structural and
controller parameters are considered
simultaneously
Cost Function
Acoustic Performance
Effort Penalty
Weight Penalty


17
Optimization Procedure Open Loop Performance
18
Optimization Procedure Features

• 1 Collocated Sensor Actuator Pair (SAP)
• Control type Velocity feedback
• Optimization concerning the thickness of the
  firewall and the feedback gain
• Cost function considering Performance, Effort
  and Weight
• MSC.Nastran used for structural calculations
• LMS.Sysnoise used for acoustic and
  vibro-acoustic calculation
• Matlab/Simulink used for state-space simulation
• LMS.Optimus used as simulation management and
  optimal search engine

19
Optimization Procedure (cont.)
LMS OPTIMUS
Structural Analysis - MSC.Nastran
thickness
Coupled Model - Sysnoise
State Space Model SDT/Matlab
gain
gain
closed-loop results
mass
COST FUNCTION
20
Optimization Procedure Two Cases
Concurrently Optimization
Discrete Extensive Search
Fixed Position
Optimizing Thickness and Feedback Gain and
Position
Optimizing Thickness and Feedback Gain
21
Optimization Results Fixed Position
Arbitrary position (node 171)
Design Table Several Discrete Data
Response Surface Model (RSM) Continuous Data
Optimization using the fitted model
22
Optimization Results
Optimization Results Discrete Extensive Search

• Fixed number of possible thicknesses0.5, 1.0,
  1.5, 2.0, 2.5, 3.0, 3.5, and 4.0mm

• Extensive Search for every possible location for
  the SAP

Optimum Solution
Best Points for each thickness
23
Optimization Results Discrete Extensive Search
Comparison of various local Optima for each
Thickness
24
Concurrent Design X Classical Design
The Optimum Solution has better performance than
all the open loop.
A lighter Active Solution can perform better than
any heavier Passive System
Classical design would lead to a thicker
firewall (2.6mm control system) However with
worse Cost Function value
25
Summary and Conclusions

• Methodology using MSC.Nastran LMS.Sysnoise to
  obtain a fully coupled model
• This fully coupled model can be reduced and
  converted to a MIMO State Space model
• Automatic communication between FEM software and
  Matlab allows the use of a Optimization Engine
  (LMS.Optimus)

• As far as the amount of design parameters
  increase, a systematic methodology is needed,
  since experience and insight may not be
  sufficient
• The reduced SS models (quick simulation) allows
  optimization procedures
• Concurrent design leads to lighter and better
  solution than conventional design

26
References
1 Herman Van der Auweraer, Sven Herold, Jan
Mohring, Leopoldo de Oliveira, Maíra da Silva and
Arnaud Deraemaeker, CAE Approach to the design
of smart structures applications Euronoise 2006
Tampere, Finland 2 Leopoldo de Oliveira, Bert
Stallaert, Wim Desmet, Jan Swevers, Paul Sas,
Optimization Strategies for Decentralized ASAC
Forum Acusticum 2005 Budapest, Hungary 3
W.Desmet, D.Vandepite, Finite Element Method in
Acoustic Seminar on Advanced Techniques in
Applied and Numerical Acoustics - ISAAC 16, 2005
Leuven, Belgim 4 J.A.Reyer, P.Y.Papalambros,
An Investigation into Modeling and Solution
Strategies for Optimal Design and Control, ASME
2000 Design Engineering Technical Conferences and
Computers and Information Engineering Conference,
Baltimore Maryland. 5 Van Brussel,H.,
Mechatronics - A Powerful Concurrent Engineering
Framework, IEEE/ASME Transactions on
Mechatronics, 1996, Vol. 1, Issue 2.