Probabilistic design with focus on blades

Blades

Blades

–

–

probabilistic design

probabilistic design

Sandia Blades Workshop 12-14 May 2008 Dick Veldkamp

Contents

ƒ Introduction: why probabilistic design?

Blade root failure probability: statistics

ƒ DNV: design for p = 10-5 _{per year (component)}

ƒ Dutch Handbook for Risk Assessment of Wind Turbines

ƒ p = 6.3×10-4 _{(expected value, wind turbine)}

Are new blades safer?

ƒ Suppose we have N years without n = 0 failures, while p_fail = p:

ƒ P₀ is “amount of luck”

Are new blades safer?

1.E+02 1.E+03 1.E+04 1.E+05 1.E+06 WT operating time [year]

Conclusion on failure statistics

ƒ Annual failure probability (lower than) p = 10-5 _{is demanded.}

ƒ It will be a long time before it can be proved with data from actual

Objectives

1. Given target failure rate, find a set of optimal partial safety factors, giving

minimum weight safe design. Factors may be for:

1. loads

2. modelling

3. material properties

2. Better insight in uncertainties in calculation methods

1. most important uncertainty?

2. improved risk assessment (absolute and relative assessment)

3. Because of time constraints, work is limited to fatigue of the structure,

Economic reasons

Cost [Euro ct/kWh] Wind Conventional

Minimum weight design

ƒ How large γ_f and γ_m must be is determined by probabilistic considerations: which failure probability is allowed?

Which failure probability is allowed?

ƒ Gene Kranz, during the (movie of the) rescue of the Apollo 13, 1969.

ƒ Failure is an option, we just don’t want it to happen very often. ƒ Dick Veldkamp, 2006.

ƒ How often may be set by:

ƒ public safety considerations: ca p = 10-5 _{per year}

Experiment: 100.000 times

1. Build a turbine

2. Measure wind load and blade strength

3. Result (failure yes/no)

2. Simulate wind load and blade strength by

Monte Carlo analysis

Example: base case

ƒ IEC II design for and IEC class II site

ƒ Wind speed: U_design = U_site = 8.5 m/s at hub height

ƒ Turbulence intensity: I_design = 18%, I_site = 16% + 2% for windfarm wakes

ƒ Wind spectrum shape: Mann’s

Γ

Γ

Coefficients of variation (influence on loads)

Parameter Edge moment Flap moment

Wind (speed and turbulence)

0.04 0.04

Aerodynamic model 0.03 0.10

FEM 0.06 0.06

Stress factor (load sequence)

0.10 0.10

Aerodynamic model

ƒ Edge moments: COV = 0.03

ƒ Flap moments: COV = 0.10

ƒ Cause: BEM is not good enough

Fatigue

ƒ Stress factor: influence of load sequence and how cumujlative fatigue damage is calculated: COV = 10% (spread on life L_10%/L_90% = 10)

Required safety factor: target p

= 10

_{per year}

Parameter known

exactly

Edge moment Flap moment

None 1.33 1.45

Aerodynamic model 1.33 1.34

FEM 1.27 1.40

Stress factor 1.18 1.33

To do

ƒ Common failure database?

ƒ Blind FEM testing

ƒ Aerodynamic model