reliability-based calibration of partial safety factors for wind turbine blades
DESCRIPTION
Reliability-based Calibration of Partial Safety Factors for Wind Turbine Blades Henrik Stensgaard Toft (1) Kim Branner (2) Peter Berring (2) John Dalsgaard Sørensen (1,2) (1) Aalborg University, Denmark (2) Risø-DTU, Denmark. Outline of Presentation. Introduction Reliability-based design - PowerPoint PPT PresentationTRANSCRIPT
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
1
Reliability-based Calibration of Partial Safety Factors for Wind Turbine Blades
Henrik Stensgaard Toft(1)
Kim Branner(2)
Peter Berring(2)
John Dalsgaard Sørensen(1,2)
(1)Aalborg University, Denmark(2)Risø-DTU, Denmark
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
2
Outline of Presentation
• Introduction
• Reliability-based design
• Modelling of uncertainties
• Illustrative example: Reliability of wind turbine blade
• Conclusions & Future work
[www.lmwindpower.com]
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
3
Wind Turbine Blades
Wind turbine blades consist normally of a main spar and an aerodynamic shell.
• Blades are typically made of glass fiber reinforced epoxy, but carbon fibers are also used.
• Optimal structural design of the blades will lead to load reduction on other major wind turbine component (e.g. tower and foundation).
[www.lmwindpower.com]
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
4
Observed Annual Failure Rates and Downtime
[ISET, 2008]
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
5
Reliability-based Design
• Limit state equation:
• Probability of failure:
• Reliability index:
Reliability Estimation based on Finite Element Analysis
Evaluation of large and nonlinear FE models is very time consuming. This
demands for a simple and fast way to estimate the reliability.
• Smart simulation methods (e.g. importance sampling).
• First or Second Order Reliability methods (FORM/SORM).
• Response-surface method
g R S
FP P R S
1FP
f R, f
S
R, SRc
Sc
R ~ ResistanceS ~ Load
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
6
Design and Limit State Equation
• Design equation according to IEC 61400-1:
• Limit state equation:
• Response surface technique:
The constants a, bi and ci are determined by evaluating the limit state equation (FE-
model) in the points for each stochastic variable Xi.
The partial safety factors can be calibrated in order to obtain a target reliability
index t for each failure mode.
,1 mat cf c
n m
G R L
x
R matg X R Q X
Partial Safety Factors (IEC 61400-1) ULS
n – Consequences of failure 1.00
m – Material properties (brittle failure) 1.30
f – Load (normal operation) 1.25
2
1 1
n n
i i i ii i
g a b x c x
x
i i i ix X f
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
7
Modelling of Uncertainties
The uncertainties can be divided into the following four groups:
• Physical uncertainty (aleatory)
• Measurement uncertainty (epistemic)
• Statistical uncertainty (epistemic)
• Model uncertainty (epistemic)
The uncertainties are modelled by stochastic variables.
[www.lmwindpower.com]
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
8
Uncertainty – Material Properties
• Stochastic models for the material properties have been modelled from micro-scale using the properties of the fiber and matrix (model uncertainty included).
• Scatter between the uncertainties in the literature is observed due to e.g. variations in the manufacturing procedure and basic material.
[Lekou & Philippidis] [DNV-OS-C501] Micro-scale modelingincl. model uncertainty
Variable Distribution COV COV Distribution Mean COV
E1 [GPa] Weibull 0.089 0.050 Lognormal 39.90 0.106
E2 [GPa] Extreme Type I 0.148 0.100 Lognormal 13.62 0.136
G12 [GPa] Gamma 0.249 0.100 Lognormal 4.23 0.129
12 [-] Weibull 0.187 0.100 Lognormal 0.290 0.107
XT [MPa] Weibull 0.151 0.050 Normal 795.7 0.138
XC [MPa] Lognormal 0.101 0.050 Normal 529.7 0.143
YT [MPa] Extreme Type I 0.150 0.100 Weibull 54.0 0.104
YC [MPa] Extreme Type I 0.135 0.150 Weibull 162.6 0.112
S [MPa] Weibull 0.181 0.100 Weibull 55.8 0.106
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
9
Uncertainty – Load-effect
• The uncertainty related to the load-effect corresponds to DLC 1.1 in IEC 61400-1 (normal operation – blade out-of-plane bending).
• The physical uncertainty related to the load-effect is estimated from statistical load extrapolation for the NREL 5MW reference wind turbine.
expdyn st aero str simQ X X X X X X L
Variable Description Distribution Mean COV
Xdyn Dynamics Lognormal 1.00 0.05
Xexp Exposure Lognormal 1.00 0.20
Xst Limited wind data Lognormal 1.00 0.10
Xaero Aerodynamics Gumbel 1.00 0.10
Xstr Load-effect (model) Lognormal 1.00 0.03
Xsim Limited simulations Normal 1.00 0.05
L Load-effect (physical) Weibull - 0.15
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
10
Uncertainty – Failure Prediction
• Failure of the blade is estimated by First Ply Failure (no redundancy and damage tolerance).
• Ply failure is estimated by the Tsai-Wu criteria.
• The uncertainty related to failure predictions is estimated from test results in the World Wide Failure Exercise.• Based on UD and MD-laminates.• Final failure - degradation of the structure is taken into account.
0 0.5 1 1.5 2 2.5 3 3.50
0.2
0.4
0.6
0.8
1
Ratio (test capacity / estimated capacity) [-]
Cum
ulat
ive
Pro
babi
lity
[-]
0 0.5 1 1.5 2 2.50
0.2
0.4
0.6
0.8
1
Ratio (test capacity / estimated capacity) [-]
Cum
ulat
ive
Pro
babi
lity
[-]
Part A: XR~LN(1.31; 0.40) Part B: XR~LN(1.14; 0.30)
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
11
Illustrative Example: Reliability Wind Turbine Blade
• Blade for a 1.5MW pitch controlled wind turbine (shortened after 25.4m).
• Geometrical nonlinear finite element analysis (mesh size 40x40mm).
• Load case: Combined edgewise and flapwise loading (normal operation).
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
12
Illustrative Example: Reliability Wind Turbine Blade
• Buckling of main spar cap at 3.4m, 9.0m, 11.6 and 13.6m.
• First ply failure (Tsai-Wu) in UD-lamina due to high strains in transverse direction (9m from blade root).
• Failure mode seems robust related to changes of the material properties.
0 5 10 15 20 250
0.2
0.4
0.6
0.8
1
Distance from Root z [m]
Tsa
i-W
u C
rite
ria
[-]
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
13
Illustrative Example: Reliability Wind Turbine Blade
The annual reliability is estimated using the response surface technique and Monte Carlo simulation / FORM.
Method / Parameters PF
Response surface, Monte Carlo, XR~LN(1.14,0.30) 2.7 3.210-3
Response surface, Monte Carlo, XR~LN(1.00,0.30) 2.6 4.810-3
Response surface, FORM, XR~LN(1.14,0.30) 3.1 0.910-3
Implicit target reliability in IEC 61400-1: =3.1 corresponding to PF=10-3.
0 5 10 15 20 25
3
4
5
6
7
8
9
Distance from Root z [m]
Ann
ual R
elia
bilit
y In
dex
[-
]
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
14
Illustrative Example: Reliability Wind Turbine Blade
The most important stochastic variables are determined from the-vectorestimated by FORM:
• Model uncertainty failure criteria XR (2=0.60).• Model uncertainty exposure Xexp (2=0.26).• The material properties has in general only a small influence.
The estimated reliability varies significantly dependent on the points which is used
for estimating the response surface. Especially the points for the load and failure
criteria are important. i i i ix X f
0 0.2 0.4 0.6 0.8 10
0.2
0.4
0.6
0.8
1
1.2
Load Level [-]
Tsa
i-W
u C
rite
ria
[-]
z = 3.44mz = 9.00mz =11.60mz =13.72m
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
15
Conclusion
• Stochastic models for the uncertainty related to the material properties, load-effect and failure prediction for wind turbine blades have been proposed.
• The reliability have of the blade have been estimated using the response surface technique based on nonlinear finite element analysis.
• The estimated reliability level is slightly lower than the target reliability in IEC 61400-1.
Future Work
• The uncertainty related to failure prediction should be studied further in order to improve the stochastic models.
• The reliability is sensitive to the points used for evaluating the response surface. This demands for a more robust way of estimating the response surface.
• Estimate the influence of system effects (redundancy) and damage tolerance.
Department of Civil Engineering Division of Structural Mechanics
EWEA 2011 – 16 March 2011 www.civil.aau.dk
16
Reliability-based Calibration of Partial Safety Factors for Wind Turbine Blades
Henrik Stensgaard Toft(1)
Kim Branner(2)
Peter Berring(2)
John Dalsgaard Sørensen(1,2)
(1)Aalborg University, Denmark(2)Risø-DTU, Denmark