Computational Design + Optimisation CDOformpig.com/pdf/formpig_arup cdo short_shea.pdf · 2013. 10....

5 July 2005

Computational Design + Optimisation CDO

Dr Kristina Shea

Arup Foresight + Innovation, London

Engineering Design Centre (EDC)

University of Cambridge Engineering Department

Dr Kristina Shea, Foresight + Innovation

Drivers for CDO

• increased competitiondesire for improved quality

• shorter design time and cost

• increased complexity

• globalizationincreased collaboration among disciplinesdistributed design teams

• increased computing power

• increased software capabilities

• increased computer fluency of designers


With clients demanding high-quality engineering solutions for more and

more complex projects,

Can CDO provide the competitive edge in meeting and exceeding increasing

demands?


• improve design innovation and quality

• improve project delivery time and cost

• aid multi-disciplinary negotiations

• enhance design understanding

• quantify design alternatives and

performances

• find feasible design alternatives

• find “optimal” or satisfactory solutions

Potential benefits of CDO

Arup Explores Computational Optimisation, March 2004


Emphasis AreasElement

Scale: 1

structures

facades building envelope optimisation (BEO)


Design Optimization

•select variables, to describe design alternatives

•select criteria, expressed in terms of the design variables, which we seek to minimize or maximize

•determine a set of constraints, expressed in terms of design variables, which must be satisfied

•determine a set of values for design variables, which minimize or maximize the objective while satisfying all the constraints

adopted from: Principles of optimal design: Modeling and Computation, Papalambros and Wilde, 1986


Optimisation Modelling + Methods

2

3

4

5

6

7

8

9

10

1 2 3 4 5 6 7 8 9Web thickness (mm)

Flan

ge th

ickn

ess

(mm

)

FeasibleDesignSpace

InfeasibleDesignSpace

CO Method 1

CO Method 2

Deflection Limit

Shea

r Lim

itDecreasing steel m

ass


optimisation method

xi+1 = xi + αisi

min/max f(x) s.t.: g(x) ≤ 0

h(x) = 0

analysis engine

design modelx = (x1, x2...xn)T

g(x) ≤ 0h(x) = 0

decision modelf(x)

start

finish

xf = (x1, x2...xn)T

xo = (x1, x2...xn)T

Optimisation Methods


optimisation method

Excel, Matlab, LSOpt, SOL2000,

own codeanalysis/evaluation

GSA, Strand7, Nastran, LSDyna, Matlab,Excel,...

design modelCAD, FEA model, Excel,

Matlab...

decision modelExcel, SOL2000, LSOpt,

Matlab, own code

Integrated Design Tools

start

finish


Steel Section Size Optimisation

• design the most cost effective solution for lateral stiffness of tall buildings

North East Tower, Hong Kong Station (88 stories, 420 m)Prof. C. M. Chan (HKUST) and Arup Hong Kong

• optimisation model

hybrid steel / concrete

minimum overall weight

minimum overall cost including floor area


Optimised Section Sizing

Arup Sydney Group (Mark Arkinstall)

optimised distribution of beam strength utilisations

0

2000

4000

6000

8000

10000

12000

14000

16000

0 5 10 15 20 25

Optimisation Iteration Number

Num

ber o

f Ove

rstre

ssed

Mem

bers

0

1000

2000

3000

4000

5000

6000

Tota

l Ste

el T

onna

ge

Overstressed MembersSteel Tonnage


“in-house” techniques • heuristic and intuitive generate-and-test methods

• helpful in identifying feasible solutions

• limited scope, unlikely to find true optimal solutions, but relatively robust

• substantial cost benefit to projects

Beijing National Stadium Leadenhall Street

Steel Section Size Selection + Automation


Shakhatar Stadiumwith Arup Sport (Martin Simpson, Forest

Flager, Tristan Simmonds, Erin Morrow)


Design Concept• Roof

SUN EXPOSURE DIAGRAM

• Steel space frame structure

• Partially translucent roofing to increase sunlight on pitch


Virtual Prototyping and Optimization

• ‘Smart’ Parts

SUPPORT STRUCTURE COMPONENT

c1c2

c3

p1 p2

p3 p4

p5p6

Catia

GSA

code checker

optimisation


Evolving Optimised Bracing Systems for Tall Buildings

• bracing system for lateral stability of a tall building (~260m)

• efficiency in bracing design driven byconnection costsmaterial costsconstruction costs

• complex structure, behaviour not well understood

• aesthetic requirements important

CDO potential benefitsimprove design performanceenhance design understanding

BEL4 (Damian Eley, Chris Neighbour) Cambridge University (Rob Baldock)


Design Optimisation Model

Minimize ∑=

=n

SSLN

1

Subject to bracingbracing FF (max)≤

beamsbeams MM (max)≤

where:N = total number of bracing elements in the structure;LS = number of bracing elements in spiral S (design variables);n = number of bracing spirals in the structure (up to 45);Fbracing = current maximum axial force occurring in any bracing element in the structure;F(max)bracing = limit on axial force in bracing elements;Mbeams = current maximum bending moment occurring in any horizontal beam in thestructure;M(max)beams = limit on axial force in bracing elements.

3x1048 possible designs!!!

Design 0 - fully braced - 718 bracing elements

front back Max abs. axial force in bracing = 6750 kNMax abs. bending moment in beams = 487 kNm

Scale: 1:1413.

Axial Force, Fx: 25000. kN/pic.cmMinimum values of env.

-0.4943 kN-964.7 kN-1929. kN-2893. kN-3857. kN-4821. kN-5786. kN-6750. kN

Case: C24 :"ULS envelope"

Design 0 - fully braced - axial force in bracing

Max abs. axial force in bracing = 6750 kNMax abs. bending moment in beams = 487 kNm

Scale: 1:1413.

Moment, Myy: 700.0 kN m/pic.cmAbsolute value of env.

486.7 kN m417.2 kN m347.7 kN m278.1 kN m208.6 kN m139.1 kN m69.53 kN m3.776E-6 kN m


Design 0 - fully braced - bending moment in beams

Max abs. axial force in bracing = 6750 kNMax abs. bending moment in beams = 487 kNm

19

1718 3637

38

1516 3233

3435

14 2930

31

27

28

4

2 3 5

1 62122

720

8

9

10 1326

11 1223

2524

spiral tips available for immediate removal

Design 1 - 337 bracing elements

front backMax abs. axial force in bracing = 8355 kNMax abs. bending moment in beams = 433 kNmMethod: basic automation

Scale: 1:1413.

Axial Force, Fx: 30000. kN/pic.cmMinimum values of env.

-478.7 kN-1604. kN-2729. kN-3854. kN-4980. kN-6105. kN-7230. kN-8355. kN


Design 1 - axial force in bracingMax abs. axial force in bracing = 8355 kNMax abs. bending moment in beams = 433 kNmMethod: basic automation

Design 1 - bending moment in beams

Scale: 1:1413.

Moment, Myy: 1000. kN m/pic.cmAbsolute value of env.

433.2 kN m371.3 kN m309.4 kN m247.6 kN m185.7 kN m123.8 kN m61.89 kN m1.921E-6 kN m


Max abs. axial force in bracing = 8355 kNMax abs. bending moment in beams = 433 kNmMethod: basic automation

evolutionary design optimisation process


front backMax abs. axial force in bracing = 8492 kNMax abs. bending moment in beams = 480 kNmMethod: deterministic pattern search


front backMax abs. axial force in bracing = 8451 kNMax abs. bending moment in beams = 500 kNmMethod: randomized pattern search


Initial Result Summary

Efficiency (analyses

per element removal)

Bracing Elements in

Final Structure

Max Axial Force in Final Structure (kN)

Max Bending Moment in Final Structure (kNm)

Design 0 - 718 6750 487

Design 1 12+ 337 8355 433

Design 2 2.5 319 8492 480

Design 3 2.4 319 8451 500


Parametric Design Study

303 bracing elements


384 bracing elements 296 bracing elements



500 kNmBending limit



7,000 kNBracing force limit




271 elements 251 elements 251 elements(684kNm, 8398kN) (744kNm, 8489kN) (726kNm, 8475kN)

Outline Proposals


CDO Benefits Achieved

• improved design performanceyes - improved structural performance

met aesthetic design goals

• enhanced design understandingyes - better understanding of structural parameters

better understanding of the minimum number of bracing elements

better understanding of problem areas in the design

Dr Kristina Shea, Foresight + InnovationAdvanced Geometry Unit (Cecil Balmond, Daniel Bosia)

1 generate spatial tiling

Weaire Phelan Danzer

Using Spatial Tilings in Design


Using Spatial Tilings for Design

2 adapt tiling for design context


Erosion Algorithm

0 Analyse structure (Rhino model -> GSA)

1 Check constraints (max # members, displacement, ...)

2 Remove elements with performance ratio < removal ratio

(force/max force or utilisation/max utilisation)

3 If no elements were removed, increase removal ratio if removal ratio < max removal ratio

Return to step 1 if removal ratio updated

If removal ratio ≥ max removal ratio, stop

4 Check for disconnected “small” structures

5 Reanalyse structure (Rhino model -> GSA)

6 Retrieve data, calculate performance ratios and sort members by increasing performance. Return to step 1.


3 erode structure driven by force, utilisation, displacement, ...

6732


3000


1518


987


496


300


What next?

CDO provides new opportunities for design generation, optimization and rationalization.

We will continue to expand our expertise in the area to meet and exceed demand for high-quality engineering solutions

on ever more complex projects.


aesthetic

spatial usage / program

structural

environmental control

lighting

energy

acousticsustainability

fire

Optimal Building Performance?

cost


Exploring pareto-optimal design archives

Computational Design + Optimisation CDOformpig.com/pdf/formpig_arup cdo short_shea.pdf · 2013. 10....

Documents

Transcript of Computational Design + Optimisation CDOformpig.com/pdf/formpig_arup cdo short_shea.pdf · 2013. 10....