2014-05-28

1

Engineering modelling uncertainty – results from an investigation of structural engineers

IRCC WorkshopMalmö, June 2014

Martin FröderbergSven Thelandersson

Research project: Conceptual design of structural systems ‐minimizing risks and uncertainties in the modern design process

Objectives:

• Increase efficiency of the conceptual evaluation in early design stages

• Limit the risk of errors and irregularities in structural engineering

2014-05-28

2014-05-28

2

Background of investigationHazards with the structural engineer in focus

Älandsfjärden, Härnösand 2008

2014-05-28

Kistagallerian, Stockholm 2008

Hälsans hus, Ystad 2012

Round-robin test

2014-05-28

2014-05-28

3

Round robin investigation:Simulation of a realistic engineering task in early design phase

• Participants performed the same task individually• 8 hours of paid work per task

Participants

2014-05-28

Requirements:• >5 years as structural engineer

• Experience from steel, concrete and timber engineering

Result:• 16 participants• Average experience level: 12 years• Located in the major cities of Sweden:

– Stockholm– Gothenburg– Malmö

2014-05-28

4

The building

2014-05-28

• 5 stories

– First floor: Public spaces

– 4 stories student apartments

• Semi precast structure

– 220mm slabs

– 200mm walls

– Concrete columns

The task

• Check dimensions

– Walls 200 mm concrete

– Floors 220 mm

• Column dimensions.

• Load transfer to ground

– Vertical loads

– Loads on stabilizing units

2014-05-28

2014-05-28

5

Results

• Dimensions OK!

• Columns – 300x300

• Loads…

2014-05-28

0

5

10

200

250

300

350

400

450

500

550

600

650

700

Frekvens

Pelardimension [mm]

Vertical loads (grid line 3)

2014-05-28

6

w3e e4r r5t t6y u8i a2s s3d d4f g6h k9l z2x x3c c4v v5b n7m m8l

Column B2 1011 601 690 828 1145 1456 810 1051 1190 1619 970 1100 1350 951 1494 1473

0

200

400

600

800

1000

1200

1400

1600

1800

Lo

ad[k

N]

Design load column B3

2014-05-28

Design load column C3

2014-05-28

w3e e4r r5t t6y u8i a2s s3d d4f g6h k9l z2x x3c c4v v5b n7m m8l

Column C2 1081 1420 690 1494 1061 1758 1500 1130 1465 1556 1184 1400 1350 1350 1494 1453

0

200

400

600

800

1000

1200

1400

1600

1800

2000

Lo

ad[k

N]

2014-05-28

7

Design load column D3

2014-05-28

w3e e4r r5t t6y u8i a2s s3d d4f g6h k9l z2x x3c c4v v5b n7m m8l

Column D2 688 476 690 448 393 808 580 488 900 940 593 1100 1350 844 815 998

0

200

400

600

800

1000

1200

1400

1600

Lo

ad[k

N]

Summary

2014-05-28

or

Safe area

2014-05-28

8

Load effects in the 3 stabilizing shear wallsMoments and shear forces

2014-05-28

Shear force

Moment

Uncertainty in the design process

Construction drawingsPrincipal drawingsConceptual design

Un

cert

ain

ty

Design code

EngineeringModelling Uncertainty

Gross errors

Inv

2014-05-28

9

Different experience

• Engineering modelling uncertainty, EMU– Interpretation of codes and available information very different

– Widely different structural models were utilised (guidance in code rules not sufficient)

2014-05-28

Probability of failure:PF1= 4,8·10-7

PF2= 4,3·10-3

Effect of EMU on safety(evaluated for foundation element)

2014-05-28

g´1(G,L,S,x,fy,Asd,1Q) = 1Q· fy·Asd – (G+L+S)·x/(2·a)

g´2(G,L,S,x,fy,Asd,2,2Q) = 2 ·2Q · fy·Asd – (G+L+S)·x/(2·a)

=4,9

=2,6

The probability offailure increases by a factor 104!

2014-05-28

10

Factors which may explain errors and variability in design work

2014-05-28

7%

• Education and experience

• Available time (for analysis and decisions)

• Communication

• Complexity (project, codes and tools)

Advanced analysis tools

2014-05-28

”The computer is the ultimate fool, and we haveelevated it to the ultimate authority. What is going to be the price?”

Franz Knoll, 1986

2014-05-28

11

2014-05-28

Hartford Civic Centre 1978

Sleipner 1991

Too coarse FEM-mesh shear force under-estimated by 45%

Top girders werefalsely modelled as being braced trans-versely degree ofutilization 852%

Two spectacular examples

Risks related to advanced analysis tools

• May encourage complex solutions and poor design. 

• Can generate results even if knowledge is lacking

• Over‐confidence in results

• Multiple choices of modelling assumptions can lead to widelydifferent results

• For many engineers work it works as a black box 

2014-05-28

2014-05-28

12

Too advanced code rules

2014-05-28

Eurocode formulas for design of a steel beam-column element. Contact with reality totally lost.

Negative effects of complex codified rules

• Prevents development of conceptual understanding

• Unfit for use in early stages of design  ”poor solutions”

• Independent  thinking not encouraged

• Range of applicability unclear

• Black box format  control of results difficult

2014-05-28

2014-05-28

13

Conclusions

• Grey area between model uncertainty and gross errors ‐EMU

• Poor standard of presentation of reported result

• EMU requires additional safety margins and goodcommunication

• Inappropriate use of modern analysis tools and code rulesseems to be part of the problem

2014-05-28