flexible life extension with innovative monitoring
TRANSCRIPT
Flexible Life Extension with Innovative Monitoring
14th March 2019
Christian Wiebe
Fatigue Design/Life Extension – Conventional Approach
Global Response Analysis
Local Stress Analysis
Stress Histogram Revised Fatigue Damage
Source: NORSOK
0.001
0.01
0.1
1
10
100
16-Apr 29-Apr 12-May 25-May 7-Jun 20-Jun 3-Jul 16-Jul 29-Jul 11-Aug
Fa
tig
ue
D
am
ag
e A
ccu
mu
lati
on
(%
)
Date
FATIGUE DAMAGE ACCUMULATION (%)
Fatigue Critical Location #1 Fatigue Critical Location #2
Fatigue Critical Location #3 Fatigue Critical Location #4
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Campaign I Campaign II
Red Limit = 10% of Accumulated Fatigue Damage
Amber Limit = 3% of Accumulated Fatigue Damage
Real-TimeFatigue Damage
Life Extension – New Approach
Measured Response
Artificial Neural Network (ANN)
Stress Time History
0.001
0.01
0.1
1
10
100
16-Apr 29-Apr 12-May 25-May 7-Jun 20-Jun 3-Jul 16-Jul 29-Jul 11-Aug
Fa
tig
ue
D
am
ag
e A
ccu
mu
lati
on
(%
)
Date
FATIGUE DAMAGE ACCUMULATION (%)
Fatigue Critical Location #1 Fatigue Critical Location #2
Fatigue Critical Location #3 Fatigue Critical Location #4
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Campaign I Campaign II
Red Limit = 10% of Accumulated Fatigue Damage
Amber Limit = 3% of Accumulated Fatigue Damage
Real-TimeFatigue Damage
Life Extension – New Approach
Measured Response
Artificial Neural Network (ANN)
Stress Time History
FlexAssure
▪ Vessel Translational & Angular Motion
▪ Riser Translational Motion
▪ Field Proven
INTEGRIStick
▪ Riser Curvature
▪ Proven Technology
Machine Learning
Machine Learning Model
Feedback Loop
Yes
Development
Calculate Error
Targets
Inputs
Feedback Loop
300
400
500
600
700
800
900
1,000
800 900 1000 1100 1200 1300 1400
Te
nsil
e A
rmo
ur
La
ye
r S
tre
ss (
MP
a)
Time (seconds)
Tullow - Jubilee Flexible Riser Eng Support
TARGET Vs OUTPUT TIMETRACES COMPARISON
Tensile Armour Layer Stress, Fatigue Seastate 4 (High Damage,
Low Error), 10, 10 Delays, 10 Hidden Layers
Target Output
Error in Maxima=0.20%Error in Minima=3.89%Error in Mean=0.76%Error in StDev=0.02%
Global Analysis
Stress Analysis
Meet Criteria
Continue Training
No
Yes Error
Global Analysis
Machine Learning
Output
Stress
300
400
500
600
700
800
900
1000
600 700 800 900 1000 1100 1200 1300 1400 1500 1600
Lo
ng
itu
din
al
Str
ess (
MP
a)
Time (seconds)
Flexible Riser Fatigue Tracker
MACHINE LEARNING - NARX/LM STRESS TIME TRACEFATIGUE DAMAGE, EXPONENT 6 - Seastate 7
Seastate 7 (ANN) Seastate 7 (FEA)
Fatigue
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5
Pre
dic
ted
Fa
tig
ue
Da
ma
ge
Ra
te (
1/
Ye
ar)
FEA Fatigue Damage (1/Year)
Flexible Riser Fatigue Tracker
MACHINE LEARNING - STRESS TIME TRACEENARX/LM, Exponent 6 - 251 Seastates, Rainflow Counting
Over-Prediction
Under-Prediction
Development
Calculate Error
Targets
Inputs
300
400
500
600
700
800
900
1,000
800 900 1000 1100 1200 1300 1400
Te
nsil
e A
rmo
ur
La
ye
r S
tre
ss (
MP
a)
Time (seconds)
Tullow - Jubilee Flexible Riser Eng Support
TARGET Vs OUTPUT TIMETRACES COMPARISON
Tensile Armour Layer Stress, Fatigue Seastate 4 (High Damage,
Low Error), 10, 10 Delays, 10 Hidden Layers
Target Output
Error in Maxima=0.20%Error in Minima=3.89%Error in Mean=0.76%Error in StDev=0.02%
Global Analysis
Stress Analysis
Meet Criteria
Continue Training
No
Yes Error
Global Analysis
Feedback Loop
Machine Learning
Output
Application
Inputs
300
400
500
600
700
800
900
1,000
800 900 1000 1100 1200 1300 1400
Te
nsil
e A
rmo
ur
La
ye
r S
tre
ss (
MP
a)
Time (seconds)
Tullow - Jubilee Flexible Riser Eng Support
TARGET Vs OUTPUT TIMETRACES COMPARISON
Tensile Armour Layer Stress, Fatigue Seastate 4 (High Damage,
Low Error), 10, 10 Delays, 10 Hidden Layers
Target Output
Error in Maxima=0.20%Error in Minima=3.89%Error in Mean=0.76%Error in StDev=0.02%
Global Analysis
Stress Analysis
Meet Criteria
Continue Training
No
Yes Deploy
0.001
0.01
0.1
1
10
100
16-Apr 29-Apr 12-May 25-May 7-Jun 20-Jun 3-Jul 16-Jul 29-Jul 11-Aug
Fa
tig
ue
D
am
ag
e A
ccu
mu
lati
on
(%
)
Date
FATIGUE DAMAGE ACCUMULATION (%)
Fatigue Critical Location #1 Fatigue Critical Location #2
Fatigue Critical Location #3 Fatigue Critical Location #4
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Campaign I Campaign II
Red Limit = 10% of Accumulated Fatigue Damage
Amber Limit = 3% of Accumulated Fatigue Damage
Calculate Error
Targets
Feedback Loop
Machine Learning
Output
Flexible Life Extension with Innovative Monitoring
0.001
0.01
0.1
1
10
100
16-Apr 29-Apr 12-May 25-May 7-Jun 20-Jun 3-Jul 16-Jul 29-Jul 11-Aug
Fa
tig
ue
D
am
ag
e A
ccu
mu
lati
on
(%
)
Date
FATIGUE DAMAGE ACCUMULATION (%)
Fatigue Critical Location #1 Fatigue Critical Location #2
Fatigue Critical Location #3 Fatigue Critical Location #4
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Lo
gg
er
Insta
lla
tio
n
Lo
gg
er
Re
trie
va
l
Campaign I Campaign II
Red Limit = 10% of Accumulated Fatigue Damage
Amber Limit = 3% of Accumulated Fatigue Damage
Source: NORSOK
Global Response Analysis
Local Stress Analysis
Stress Histogram Revised
Fatigue DamageReal-Time
Fatigue DamageMeasured Response
Artificial Neural Network (ANN)
Stress Time History
Flexible Riser Fatigue
NARX/Levenberg-Marquardt
▪ Typically used for time series
▪ High accuracy with minimal training datasets
Selected Training Seastates
0%
10%
20%
30%
40%
50%
60%
211
205
106
70
216
231
223
250
188
88
74
123 2
165
174
238
50
208
22
62
44
193
236
138
40
96
23
148
93
46
94
27
34
33
58
185
Co
ve
rag
e (
%)
Seastate Number
Tullow - Jubilee Flexible Riser Eng Support
SELECTION OF TRAINING SEASTATES
All Seastates Training Seastates
FlexAssure Case Study
▪ OSX-3
▪ 26 – 14 – 5
▪ Real Time
INTEGRIStick Case Study
▪ Motion & Strain
▪ Fatigue
▪ Unexpected
Presentation Title
01 Write revision description here DD Mmm YY X Yyyyyy X Yyyyyy X Yyyyyy
Rev Description Date Author Checked Approved
Document No: Client Ref:
ppppp-TTT-nnnn ppppp-ttt-nnnn-xxx-zzzzz
HARD COPIES ARE UNCONTROLLEDControlled copies of this document are only issued in the PDF format that includes digital signatures of the author, checker and approver. These signatures canbe viewed in Adobe Acrobat. The document will be digitally marked if it has been altered since signing. In order for recipients of documents to verify digitalsignatures, digital signature certificates may be requested from 2H.