Sensitivity Analysis - Dynardo GmbH
Sensitivity Analysis - Dynardo GmbH Sensitivity Analysis - Dynardo GmbH
Optimization and Robustness Evaluation of an Axial Turbine using Fluid-Structure Interaction Johannes Einzinger ANSYS Continental Europe johannes.einzinger@ansys.com Dirk Roos & Veit Bayer DYNARDO GmbH dirk.roos@dynardo.de © 2009 ANSYS, Inc. All rights reserved. 1 ANSYS, Inc. Proprietary
- Page 2 and 3: Outline • Parametric Process Inte
- Page 4 and 5: Workbench Platform & optiSLang Proc
- Page 6 and 7: Parameterization of the Workflow Pa
- Page 8 and 9: Parameterization of the Geometry 1
- Page 10 and 11: Workbench Interface to optiSLang ©
- Page 12 and 13: Parameter Attributes Input Paramete
- Page 14 and 15: Parameter Physics, Mechanic ΔT rBl
- Page 16 and 17: Sensitivity Analysis 1. Scanning th
- Page 18 and 19: Variation of Design Space initial
- Page 20 and 21: Coefficient of Importance, CoI Deac
- Page 22 and 23: Meta-Model of Prognosis, MoP Respon
- Page 24 and 25: Objective Function Target Function
- Page 26 and 27: Conclusion Sensitivity Analysis •
- Page 28 and 29: Design Optimization Gradient-Based
- Page 30 and 31: Outlook Strategy Sensitivity Analys
- Page 32 and 33: Evolutionary Algorithms (EA) • Hi
- Page 34 and 35: Design variable 2 Design variable 2
- Page 36 and 37: Adaptive Response Surface (ARSM) be
- Page 38 and 39: 1,8 1,6 1,4 1,2 1 0,8 0,6 0,4 0,2 0
- Page 40 and 41: Initial vs. Optimized Design • n
- Page 42 and 43: Robustness Evaluation 1. Scanning t
- Page 44 and 45: Correlation, CoI, Histogram Guide v
- Page 46 and 47: Evaluation of Histogram Limit for V
- Page 48 and 49: Conclusion Robustness Evaluation
- Page 50: Summary • Workbench supports full
Optimization and<br />
Robustness Evaluation<br />
of an Axial Turbine using<br />
Fluid-Structure<br />
Interaction<br />
Johannes Einzinger<br />
ANSYS Continental Europe<br />
johannes.einzinger@ansys.com<br />
Dirk Roos & Veit Bayer<br />
DYNARDO <strong>GmbH</strong><br />
dirk.roos@dynardo.de<br />
© 2009 ANSYS, Inc. All rights reserved. 1 ANSYS, Inc. Proprietary
Outline<br />
• Parametric Process Integration<br />
• <strong>Sensitivity</strong> <strong>Analysis</strong><br />
• Design Optimization<br />
– Evolutionary Algorithm<br />
– Adaptive Response Surface Method<br />
• Robustness Evaluation<br />
• Outlook<br />
– Random Fields<br />
– Design for Six Sigma (Reliability <strong>Analysis</strong>)<br />
© 2009 ANSYS, Inc. All rights reserved. 2 ANSYS, Inc. Proprietary
Motivation<br />
Power Plant 1000 MW<br />
Efficiency 50 %<br />
Increase of 1% +20 MW<br />
=Electricity for 120 000<br />
Inhabitants<br />
© 2009 ANSYS, Inc. All rights reserved. 3 ANSYS, Inc. Proprietary
Workbench Platform & optiSLang<br />
Process integration<br />
CAD / PDM<br />
ANSYS Workbench<br />
Structural Mechanics - Fluid Dynamics - Heat Transfer - Electromagnetic<br />
A Multi-Physics Design and <strong>Analysis</strong> System<br />
<strong>Sensitivity</strong> Optimization Robustness Reliability Robust Design<br />
© 2009 ANSYS, Inc. All rights reserved. 4 ANSYS, Inc. Proprietary
Parameterization<br />
Process & Geometry<br />
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
Design Optimization<br />
Robustness Evaluation<br />
© 2009 ANSYS, Inc. All rights reserved. 5 ANSYS, Inc. Proprietary
Parameterization of the Workflow<br />
Parametric Workflow for<br />
Multi Physic Application<br />
• Extended solver support<br />
• Embedded analysis tool<br />
• Automatic user interaction<br />
© 2009 ANSYS, Inc. All rights reserved. 6 ANSYS, Inc. Proprietary
Parameterization of the Geometry<br />
Define Parameter<br />
1. Rotation Angle, Guide Vane<br />
2. Rot. Angle, Shroud Profile<br />
3. Rot Angle, Hub Profile<br />
© 2009 ANSYS, Inc. All rights reserved. 8 ANSYS, Inc. Proprietary<br />
1<br />
2<br />
3
Parameterization of the Geometry<br />
1 2<br />
1. CHT Model: Multi Body Part<br />
2. CSD Model: Single Part<br />
3. Add Blend for CSD, as Parameter<br />
© 2009 ANSYS, Inc. All rights reserved. 9 ANSYS, Inc. Proprietary<br />
3
Parameter Manager<br />
Parameter List<br />
Input Parameter = 15<br />
Output Parameter = 9<br />
List of Design Points<br />
………….<br />
Linked Parameter,<br />
by Expressions, for<br />
parameter restrictions or<br />
further output<br />
© 2009 ANSYS, Inc. All rights reserved. 10 ANSYS, Inc. Proprietary
Workbench Interface to optiSLang<br />
© 2009 ANSYS, Inc. All rights reserved. 11 ANSYS, Inc. Proprietary
Workbench Interface to optiSLang<br />
© 2009 ANSYS, Inc. All rights reserved. 12 ANSYS, Inc. Proprietary
Parameter Attributes<br />
Input Parameter Parameter Name Initial Value Type<br />
Blade Angels α GV,α Hub,α Shroud 0°, 0°, 0° deterministic<br />
Rotational Velocity of Rotor Ω -2094 [rad/s] deterministic<br />
Rotor Blend Radius r Blend 1 [mm] deterministic<br />
Total Temperature Inlet T t,Inlet 1000 [K] deterministic<br />
Total Pressure Inlet p t,Inlet 400 [kPa] deterministic<br />
Pressure Outlet p out 187 [kPa] stochastic<br />
All Material Properties - - stochastic<br />
Output Parameter Parameter Name Initial Value Target<br />
Total Temperature Ratio Θ T=T t,Inlet/T t,Outlet 1.115 -<br />
Total Pressure Ratio Π p=p t,Inlet/p t,Outlet 1.673 -<br />
Torque/Power at Rotor M P, P -577 [Nm], 1.21 [MW] maximize<br />
Mass Flow Rate m 11.56 [kg/s] -<br />
Isentropic Efficiency η 71.64 [%] maximize<br />
Maximal v. Mises Stress σ max 218.6 [MPa] below limit<br />
© 2009 ANSYS, Inc. All rights reserved. 13 ANSYS, Inc. Proprietary
Parameter Physics, Fluid Flow<br />
c0 r<br />
( Π , )<br />
m& = f α<br />
p<br />
GV<br />
( )<br />
η = f Θ , Π<br />
T<br />
p<br />
w r<br />
α GV<br />
c1 r<br />
α Hub /<br />
( Ω)<br />
Shroud<br />
© 2009 ANSYS, Inc. All rights reserved. 14 ANSYS, Inc. Proprietary<br />
1<br />
u r<br />
1<br />
= f<br />
P<br />
w r<br />
c2 r<br />
P, M ~ m&<br />
Δ ⋅<br />
⋅<br />
1−2<br />
2<br />
u 2<br />
r<br />
( u c )<br />
u
Parameter Physics, Mechanic<br />
ΔT<br />
rBlend<br />
Δumax<br />
σ max<br />
© 2009 ANSYS, Inc. All rights reserved. 15 ANSYS, Inc. Proprietary
Parameterization<br />
Process & Geometry<br />
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
Design Optimization<br />
Robustness Evaluation<br />
© 2009 ANSYS, Inc. All rights reserved. 16 ANSYS, Inc. Proprietary
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
1. Scanning the Design Space with<br />
optimized LHS, variation and<br />
correlation are investigated<br />
2. Identification of important variables<br />
• Check Variation of Design Space<br />
• Check Coefficient of Importance<br />
© 2009 ANSYS, Inc. All rights reserved. 17 ANSYS, Inc. Proprietary
Latin Hypercube and Confidences<br />
• n = 7 design variables<br />
• N = 40 design evaluations (4 failed)<br />
• Confidence levels are quite acceptable<br />
© 2009 ANSYS, Inc. All rights reserved. 18 ANSYS, Inc. Proprietary
Variation of Design Space<br />
initial<br />
© 2009 ANSYS, Inc. All rights reserved. 19 ANSYS, Inc. Proprietary
Coefficient of Importance, CoI<br />
CoI=94%>80% CoI=91%>80%<br />
© 2009 ANSYS, Inc. All rights reserved. 20 ANSYS, Inc. Proprietary
Coefficient of Importance, CoI<br />
Deactivation<br />
of outliers<br />
50% variance of the stress variation can be<br />
explained by the given n = 7 design variables<br />
© 2009 ANSYS, Inc. All rights reserved. 21 ANSYS, Inc. Proprietary
Meta-Model of Prognosis, MoP<br />
Response Surface<br />
Output: Power<br />
Meta-Model<br />
Output: Power<br />
© 2009 ANSYS, Inc. All rights reserved. 22 ANSYS, Inc. Proprietary
Meta-Model of Prognosis, MoP<br />
Response Surface<br />
Output: Efficiency<br />
Meta-Model<br />
Output: Efficiency<br />
© 2009 ANSYS, Inc. All rights reserved. 23 ANSYS, Inc. Proprietary
Coefficient of Importance, CoI vs.<br />
Coefficient of Prognosis, CoP<br />
© 2009 ANSYS, Inc. All rights reserved. 26 ANSYS, Inc. Proprietary
Objective Function<br />
Target Function for Optimization:<br />
⎛ −<br />
⋅<br />
1 P ⎞<br />
= ( 1−<br />
) + ⎜⎜ 1 ⎟⎟ min;<br />
4 ⎝ 2[<br />
MW]<br />
⎠<br />
f η = σ < σ<br />
Target<br />
seqv<br />
Limit<br />
© 2009 ANSYS, Inc. All rights reserved. 27 ANSYS, Inc. Proprietary<br />
!
Anthill Plots Objective<br />
initial<br />
Good Design<br />
Points<br />
© 2009 ANSYS, Inc. All rights reserved. 28 ANSYS, Inc. Proprietary
Conclusion <strong>Sensitivity</strong> <strong>Analysis</strong><br />
• Is <strong>Sensitivity</strong> reliable, CoI?<br />
• Is <strong>Sensitivity</strong> reliable, CoP?<br />
• Is <strong>Sensitivity</strong> plausible, physics?<br />
© 2009 ANSYS, Inc. All rights reserved. 29 ANSYS, Inc. Proprietary
Parameterization<br />
Process & Geometry<br />
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
Design Optimization<br />
Robustness Evaluation<br />
© 2009 ANSYS, Inc. All rights reserved. 30 ANSYS, Inc. Proprietary
Design Optimization<br />
Gradient-Based<br />
Algorithms<br />
Response<br />
Surface Method<br />
Evolutionary<br />
Algorithm<br />
Local<br />
Adaptive RSM<br />
Pareto<br />
Optimization<br />
Global<br />
Adaptive RSM<br />
© 2009 ANSYS, Inc. All rights reserved. 31 ANSYS, Inc. Proprietary
Optimization Strategy<br />
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
• Shows Potential<br />
• Indicates multiple local<br />
optima<br />
• No parameter reduction<br />
Strategy:<br />
• Global search, EA<br />
• Start design(s) from SA<br />
• Local improvement, ARSM<br />
initial<br />
© 2009 ANSYS, Inc. All rights reserved. 33 ANSYS, Inc. Proprietary
Outlook Strategy<br />
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
• Shows Potential<br />
• Indicates global optimum<br />
• Parameter reduction<br />
Strategy:<br />
• Pre-optimization in sub space,<br />
ARSM<br />
• Local improvement, EA (full space)<br />
• Start design(s) from ARSM<br />
© 2009 ANSYS, Inc. All rights reserved. 34 ANSYS, Inc. Proprietary
Evolutionary Algorithms (EA)<br />
Optimization in Nature:<br />
- Survival of the fittest<br />
- Evolution due to mutation, recombination and selection<br />
Evolution Algorithms [EA]<br />
Genetic Algorithms [GA]<br />
© 2009 ANSYS, Inc. All rights reserved. 35 ANSYS, Inc. Proprietary
Evolutionary Algorithms (EA)<br />
• History of the Evolutionary Algorithm<br />
© 2009 ANSYS, Inc. All rights reserved. 36 ANSYS, Inc. Proprietary
Evolutionary Algorithms (EA)<br />
best design 257<br />
start population<br />
initial<br />
• Due the non-convex<br />
behavior of the<br />
efficiency and<br />
nonlinear power<br />
function a global<br />
search strategy<br />
using genetic<br />
algorithm is<br />
recommended<br />
© 2009 ANSYS, Inc. All rights reserved. 37 ANSYS, Inc. Proprietary
Design variable 2<br />
Design variable 2<br />
Adaptive Response Surface<br />
Methods (Local)<br />
Design variable 1<br />
Design variable 1<br />
objective<br />
objective<br />
1. Iteration<br />
© 2009 ANSYS, Inc. All rights reserved. 38 ANSYS, Inc. Proprietary<br />
objective<br />
5. Iteration<br />
3. Iteration
Adaptive Response Surface (ARSM)<br />
• The ARSM does not provide differentiable and smooth<br />
problems; very efficient for n < 15 design parameters<br />
• Starting solution is based on the best design of the EA<br />
• The design space is reduced to 20% around start solution<br />
© 2009 ANSYS, Inc. All rights reserved. 39 ANSYS, Inc. Proprietary
Adaptive Response Surface (ARSM)<br />
best design ARSM<br />
best design EA<br />
start population<br />
© 2009 ANSYS, Inc. All rights reserved. 40 ANSYS, Inc. Proprietary<br />
initial
Initial vs. Optimized Design<br />
Input Initial<br />
Design<br />
Optimized<br />
α Hub 0 -0.34<br />
α Shroud 0 -0.18<br />
Ω [rev/s] -335 -365<br />
α Guide Vane 0 -9.68<br />
Output Initial Design Optimized<br />
T t Ratio 1.116 1.151<br />
p t Ratio 1.674 1.848<br />
η [%] 71.65 81.54<br />
Power [MW] 1.208 1.329<br />
© 2009 ANSYS, Inc. All rights reserved. 41 ANSYS, Inc. Proprietary
1,8<br />
1,6<br />
1,4<br />
1,2<br />
1<br />
0,8<br />
0,6<br />
0,4<br />
0,2<br />
0<br />
Initial vs. Optimized Design<br />
Initial<br />
Optimized<br />
Initial<br />
Optimized<br />
Span=0.95<br />
Span=0.50<br />
Isentropic Mach Number<br />
Initial vs. Optimized<br />
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1<br />
© 2009 ANSYS, Inc. All rights reserved. 42 ANSYS, Inc. Proprietary
Initial vs. Optimized Design<br />
Isotropic Mach Number<br />
Initial Rotor<br />
Optimized<br />
© 2009 ANSYS, Inc. All rights reserved. 43 ANSYS, Inc. Proprietary
Initial vs. Optimized Design<br />
• n = 7 design variables<br />
• N = 76 + 257 + 84 = 417 design evaluations<br />
(SA + EA + ARSM)<br />
• How robust is the initial design?<br />
• How robust is the optimized design?<br />
• How reliable is the optimized design?<br />
• How large is the influence of surface uncertainties?<br />
Output Initial SA EA ARSM<br />
Objective 1.0766 0.90034 0.86841 0.86259<br />
η [%] 71.65 80.60 81.26 81.54<br />
Power [MW] 1.208 1.304 1.343 1.329<br />
© 2009 ANSYS, Inc. All rights reserved. 44 ANSYS, Inc. Proprietary
Parameterization<br />
Process & Geometry<br />
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
Design Optimization<br />
Robustness Evaluation<br />
© 2009 ANSYS, Inc. All rights reserved. 45 ANSYS, Inc. Proprietary
Robustness Evaluation<br />
1. Scanning the Robust Space with<br />
optimized LHS, variation and<br />
correlation are investigated<br />
2. Identification of important variables<br />
• Check Variation of Robust Space<br />
• Check Histogram, limits, probabilities<br />
• Check Coefficient of Importance<br />
© 2009 ANSYS, Inc. All rights reserved. 46 ANSYS, Inc. Proprietary
Variation of Robust Space<br />
• n = 15 random parameters<br />
• N = 50 design evaluations<br />
• Initial vs. optimized design<br />
© 2009 ANSYS, Inc. All rights reserved. 48 ANSYS, Inc. Proprietary
Correlation, CoI, Histogram<br />
Guide vane angle<br />
vs. mass flow rate<br />
© 2009 ANSYS, Inc. All rights reserved. 49 ANSYS, Inc. Proprietary
Define Objective Limits<br />
• Limit state conditions<br />
• Random parameters<br />
© 2009 ANSYS, Inc. All rights reserved. 50 ANSYS, Inc. Proprietary
Evaluation of Histogram<br />
Limit for Variable<br />
© 2009 ANSYS, Inc. All rights reserved. 51 ANSYS, Inc. Proprietary
Evaluation of Histogram<br />
Non-robust<br />
initial design<br />
Robust<br />
optimized<br />
design up to a<br />
sigma level of<br />
4.5<br />
© 2009 ANSYS, Inc. All rights reserved. 52 ANSYS, Inc. Proprietary
Conclusion Robustness Evaluation<br />
• Is there a problem?<br />
• Can we explain, CoI?<br />
• Who is responsible?<br />
CoI>80%<br />
Allowed Limit<br />
© 2009 ANSYS, Inc. All rights reserved. 53 ANSYS, Inc. Proprietary
Parameterization<br />
Process & Geometry<br />
<strong>Sensitivity</strong> <strong>Analysis</strong><br />
Design Optimization<br />
Robustness Evaluation<br />
Summary<br />
© 2009 ANSYS, Inc. All rights reserved. 65 ANSYS, Inc. Proprietary
Summary<br />
• Workbench supports full Workflow<br />
– Geometry, Meshing, Simulation, Post-Processing<br />
• Multi Physics support<br />
• Parametric Workflow management<br />
• Automatic and embedded solution procedure<br />
• <strong>Sensitivity</strong> <strong>Analysis</strong><br />
• Design Optimization<br />
• Robustness Evaluation<br />
© 2009 ANSYS, Inc. All rights reserved. 66 ANSYS, Inc. Proprietary