Upload
lamthuan
View
226
Download
2
Embed Size (px)
Citation preview
Monte Carlo Simulation for Reliability and Availability analyses
2
Monte Carlo Simulation: Why?
• Computation of system reliability and availability for industrial systems characterized by:
• many components
• multistate components
• complex architectures and production logics
• complex maintenance policy
• The use of analytical methods is unfeasible
• Monte Carlo Simulation
A real example of Monte Carlo Simulation
PRODUCTION AVAILABILITY EVALUATION OF AN
OFFSHORE INSTALLATIONZio, E., Baraldi, P., Patelli E. Assessment of the availability of an offshore installation by Monte Carlo simulation. International
Journal of Pressure Vessels and Piping 83 (2006) 312–320
4System description:
basic scheme
TC
TEG
50%
50%
Wells Separation
Gas Export
3.0 mSm3/d, 60b
Oil export
TC
Flare
Oil Trt°
Water Inj.Sea
Wat. Trt°
60 b
5System description:
gas-lift
5
First loop
Gas-lift pressure Production of the Well
100 100%
60 80%
0 60%
60 b
6System description:
fuel gas generation and distribution
Second loop
7System description:
electricity power production and distribution
8
2.2 MSm3/d
2.2 MSm3/d
4.4 MSm3/d
Production Gas : 5 MSm
3/d
Oil : 26500 Sm3/ d
Water : 8000 Sm3/d
4.4 MSm3/d
23300 Sm3/d
7000 Sm3/d
TC
TEG
Fuel Gas 25 b
50%
50%
Gas Lift 100b
Wells Separation
50%
50%
Gas Lift 60b
Gas Export 3.0 MSm
3/d, 60b
Oil export
TC
Flare
EC
Oil Trt °
TG
TG
Water Inj . Sea Wat . Trt °
13 MW
13 MW
7 MW
7 MW
6 MW
6 MW
0.1
MSm3
0.1
MSm3
0.1
MSm3
0.1
MSm3
1
MSm3/d
The offshore production plant
9
State 0 = as good as new
State 1 = degraded (no function lost, greater failure rate value)
State 2 = critical (function is lost)
2
l12
l02
m20
0 1
l01
m10
TC TG
l01 0.89 · 10-3 h-1 0.67 · 10-3 h-1
l02 0.77 · 10-3 h-1 0.74 · 10-3 h-1
l12 1.86 · 10-3 h-1 2.12 · 10-3 h-1
m10 0.033 h-1 0.032 h-1
m20 0.048 h-1 0.038 h-1
Component failures and repairs:
TCs and TGs
10Component failures and repairs:
EC and TEG
0 2
l
m
State 0 = as good as new
State 2 = critical (function is lost)
EC TEG
l 0.17 · 10-3 h-
1
5.7 · 10-5 h-1
m 0.032 h-1 0.333 h-1
11
2.2 MSm3/d
2.2 MSm3/d
4.4 MSm3/d
Production Gas : 5 MSm
3/d
Oil : 26500 Sm3/ d
Water : 8000 Sm3/d
4.4 MSm3/d
23300 Sm3/d
7000 Sm3/d
TC
TEG
Fuel Gas 25 b
50%
50%
Gas Lift 100b
Wells Separation
50%
50%
Gas Lift 60b
Gas Export 3.0 MSm
3/d, 60b
Oil export
TC
Flare
EC
Oil Trt °
TG
TG
Water Inj . Sea Wat . Trt °
13 MW
13 MW
7 MW
6 MW
0.1
MSm3
0.1
MSm3
0.1
MSm3
0.1
MSm3
1
MSm3/d
Production priority: example
12
Production priority
When a failure occurs, the system is reconfigured tominimise (in order):
• the impact on the export oil production
• the impact on export gas production
➢ The impact on water injection does not matter
13
2.2 MSm3/d
2.2 MSm3/d
4.4 MSm3/d
Production Gas : 5 MSm
3/d
Oil : 26500 Sm3/ d
Water : 8000 Sm3/d
4.4 MSm3/d
23300 Sm3/d
7000 Sm3/d
TC
TEG
Fuel Gas 25 b
50%
50%
Gas Lift 100b
Wells Separation
50%
50%
Gas Lift 60b
Gas Export 3.0 MSm
3/d, 60b
Oil export
TC
Flare
EC
Oil Trt °
TG
TG
Water Inj . Sea Wat . Trt °
13 MW
13 MW
7 MW
6 MW
0.1
MSm3
0.1
MSm3
0.1
MSm3
0.1
MSm3
1
MSm3/d
Production priority: example
14
Maintenance policy: reparation
Only 1 repair team
Two or more components are failed at the same time
Priority levels of failures:1. Failures leading to total loss of export oil (both TG’s or both TC’s
or TEG)
2. Failures leading to partial loss of export oil (single TG or EC)
3. Failures leading to no loss of export oil (single TC failure)
15Maintenance policy: preventive
maintenance
➢ Only 1 preventive maintenance team
➢ The preventive maintenance takes place only if the system is in perfect state of operation
Type of maintenance Frequency [hours] Duration [hours]
Type 1 2160 (90 days) 4
Type 2 8760 (1 year) 120 (5 days) Turbo-Generator and
Turbo-Compressors Type 3 43800 (5 years) 672 (4 weeks)
Electro Compressor Type 4 2666 113
16
Problem Statement
• Compute the system istantaneous and average availability
• The use of analytical methods is unfeasible
• Monte Carlo simulation
Monte Carlo Simulation
18
Buffon
Gosset (coefficient of the t-Student)
Fermi, von Neumann, Ulam
Neutron transport
System Transport (RAM,…)
Kelvin (kinetic theory of gas
- multidimensional integrals)
The History of Monte Carlo Simulation
1707-88
1908
1824-1907
1950’s
1930-40’s
1990’s
Random
sampling of
numbers
(Manhattan project,
interaction neutron-matter)
(design of shielding in
nuclear devices -
6 dimension-integral)
19
Simulation of system transport
Simulation for reliability/availability analysis of a component
Example
Contents
Sampling Random Numbers
SAMPLING RANDOM NUMBERS
21
rrRPrUR :cdf
1 :pdf dr
rdUru R
R
0
Sampling (pseudo) Random Numbers
Uniform Distribution
0𝒖𝑹(𝒓)
𝑼𝑹(𝒓)
22Sampling Random Numbers from an
Uniform Distribution
1 million of random numbers
In the past: The Rand Book (1955)
23Sampling Random Numbers from an
Uniform Distribution
Today:
𝑹~𝑼[𝟎, 𝟏)
24Sampling Random Numbers from a
Generic Distribution
𝑥
𝐹𝑋(𝑥)
1
25Sampling Random Numbers from a
Generic Distribution
𝑥
𝐹𝑋(𝑥)
1
𝑅
𝑋 = 𝐹−1(𝑅)
• Sample R from UR(r)
• RFX X
1
26
Sample R from UR(r) and find X:
RFX X
1
Question: which distribution does X obey?
xRFPxXP X 1
1
1 r 0
rUrR RPr xFX
R X x
Sampling Random Numbers from a
Generic Distribution
F
R
27
Sample R from UR(r) and find X:
RFX X
1
Question: which distribution does X obey?
xRFPxXP X 1
1
1 r 0
rUrR RPr xFX
R X x
Sampling (pseudo) Random Numbers
Generic Distribution
F
xFRP X
R
28
Sample R from UR(r) and find X:
RFX X
1
Question: which distribution does X obey?
xRFPxXP X 1
1
1 r 0
rUrR RPr xFX
R X x
Sampling (pseudo) Random Numbers
Generic Distribution
F
xFxFRP XX
R
29
Sample R from UR(r) and find X:
RFX X
1
Question: which distribution does X obey?
Application of the operator Fx to the argument of P above yields
Summary: From an R UR(r) we obtain an X FX(x)
xRFPxXP X 1
xFxFRPxXP XX
1
1 r 0
rUrR RPr xFX
R X x
Sampling (pseudo) Random Numbers
Generic Distribution
R
30
Sampling from a Bernoulli distribution
• 𝑋 = ൝1 ′ℎ𝑒𝑎𝑙𝑡ℎ𝑦′
0 ′𝑓𝑎𝑖𝑙𝑒𝑑 ′with P 𝑋 = 0 = 𝑝0 = 0.1
𝑥
𝑃𝑋(𝑥)
0 1
0.9
𝑝𝑜 =0.1
𝑥
𝐹𝑋(𝑥)
0 1
1
𝑝𝑜=0.1
31
Sampling from a Bernoulli distribution
• 𝑋 = ൝1 ′ℎ𝑒𝑎𝑙𝑡ℎ𝑦′
0 ′𝑓𝑎𝑖𝑙𝑒𝑑 ′with P 𝑋 = 0 = 𝑝0 = 0.1
𝑥
𝑃𝑋(𝑥)
0 1
0.9
𝑝𝑜 =0.1
𝑥
𝐹𝑋(𝑥)
0 1
1
𝑝𝑜=0.1
• Sample R from UR(r)
• If 𝑅 ≤ 𝑝0 → 𝑋 = 0• If 𝑅 > 𝑝0 → 𝑋 = 1
0 𝒑𝟎 1
𝑅
Does it work?
𝑃 𝑋 = 0 = 𝑃 𝑅 ≤ 𝑝0 = 𝑝0𝑃 𝑋 = 1 = 𝑃 𝑅 > 𝑝0 = 1 − 𝑝0
𝑅
=
32
Ω = 𝑥0, 𝑥1, … , 𝑥𝑘
𝑃 𝑋 = 𝑥𝑖 = 𝑓𝑖
𝐹𝑖 = 𝑃 𝑋 ≤ 𝑥𝑖 =
𝑗=0
𝑖
𝑓𝑗
Sampling from discrete distributions
33
Ω = 𝑥0, 𝑥1, … , 𝑥𝑘 , …
𝑃 𝑋 = 𝑥𝑘 = 𝑓𝑘
𝐹𝑘 = 𝑃 𝑋 ≤ 𝑥𝑘 =
𝑖=0
𝑘
𝑓𝑘
Sampling procedure:
• Sample an
• Find 𝑖 such that 𝐹𝑖−1 < 𝑅 ≤ 𝐹𝑖
Graphically:
Why does it work?
𝑃 𝑋 = 𝑥𝑖 = 𝑃 𝐹𝑖−1 < 𝑅 ≤ 𝐹𝑖 = 𝐹𝑖 − 𝐹𝑖−1 = 𝑓𝑖
~ [0,1)R U
Sampling from discrete distributions
34Exercise 1: Sampling from
the Exponential Distribution
Probability density function:
Expected value and variance:
00
0)(
t
tetf t
T
ll
2
1][
1][
l
l
TVar
TE
t
fT(t)
( ) t
Tf t e ll
• Find the procedure to sample numbers from an exponential distribution.
Assume to have available a generator of random numbers from an uniform
distribution in the interval [0,1)
35
Arc number i Failure
probability Pi
1 0.050
2 0.025
3 0.050
4 0.020
5 0.075
Failure probability estimation: example
• 1- Calculate the analytic solution for the failure probability of the network, i.e., the probability of no connection between nodes S and T
• 2- Repeat the calculation with Monte Carlo simulation
36
Analytic solution
36
43
1
[ 1] [ ] 3.07 10T j
j
P X P M
3
1 1 4
3
2 2 5
4
3 1 3 5
5
4 2 3 4
[ ] 0.05 0.02 1 10
[ ] 0.025 0.075 1.875 10
[ ] 0.05 0.05 0.075 1.875 10
[ ] 0.025 0.05 0.02 7.5 10
P M p p
P M p p
P M p p p
P M p p p
1) Minimal Failure Configuration (minimal cut set):
M1={1,4}
M2={2,5}
M3={1,3,5}
M4={2,3,4}
2) Network failure probability:
rare event approximation
37
The state of arc i can be sampled by applying the inverse transform method to the set of discrete probabilities of the mutually exclusive and exhaustive states
•We calculate the arrival state for all the arcs of the newtork
•As a result of these transitions, if the system fails we add 1 to Nf
•The trial simulation then proceeds until we collect M trials.
0 1
R
S l
lB
B
1
31
l
lB
B
1
21
SRU[0,1)
,1i ip p
ip 1 ip
Monte Carlo simulation
𝑷𝑴𝑪 =𝑵𝒇
𝑴
38
• N number of trials
• NF number of trials corresponding to realization of a minimal cut set
• The failure probability is equal to NF / N
clear all;
%arc failure probabilities
p=[0.05,0.025,0.05,0.02,0.075];
nf=0;
M=1e6; %number of MC simulations
for i=1:M
%sampling of arcs fault events
for j=1:5 % simulation of arc states
r=rand
if (r<=p(j))
s(1,j)=1; % arc j is failed
otherwise
s(1,j)=0; % arc j is working
end
end
(is the network faulty?)
fault=s(i,1)*s(i,4)+s(i,2)*s(i,5)+s(i,1)*s(i,3)*s(i,5)+s(i,2)*s(i,3)*s(i,4);
if fault >= 1
nf=nf+1;
end
end
pf=nf/M; %system failure probability
For M=106, we obtain3[ 1] 3.04 10TP X
Monte Carlo simulation
SIMULATION OF SYSTEM TRANSPORT
40Monte Carlo simulation for system
reliability
• PLANT = system of Nc suitably connected components.
• COMPONENT = a subsystem of the plant (pump, valve,...) which maystay in different exclusive (multi)states (nominal, failed, stand-by,... ).Stochastic transitions from state-to-state occur at stochastic times.
• STATE of the PLANT at t = the set of the states in which the Nc
components stay at t. The states of the plant are labeled by a scalarwhich enumerates all the possible combinations of all the componentstates.
• PLANT TRANSITION = when any one of the plant componentsperforms a state transition we say that the plant has performed atransition. The time at which the plant performs the n-th transition iscalled tn and the plant state thereby entered is called kn.
• PLANT LIFE = stochastic process.
41
Random walk = realization of the system life generated by the underlying
state-transition stochastic process.
Plant life: random walk
Phase space
42
Phase Space
43
Example: System Reliability Estimation
• Divide the mission time, 𝑇𝑀, in bins and associate a counter (of the system failure) to each bin:
𝐶𝐹 1 , 𝐶𝐹 2 ,… , 𝐶𝐹 𝜏 , …
• Initialize each counter to 0:𝐶𝐹 1 = 0, 𝐶𝐹 2 ,= 0,… , 𝐶𝐹 𝜏 = 0,…
𝑇𝑀0
𝐶𝐹 1 = 0
𝐶𝐹 𝟐 = 0𝐶𝐹 𝝉 = 0
bin
44
Example: System Reliability Estimation
• Divide the mission time, 𝑇𝑀, in bins and associate a counter (of the system failure) to each bin:
𝐶𝐹 1 , 𝐶𝐹 2 ,… , 𝐶𝐹 𝜏 , …
• Initialize each counter to 0:𝐶𝐹 1 = 0, 𝐶𝐹 2 ,= 0,… , 𝐶𝐹 𝜏 = 0,…
• MC simulation of the first system life𝑇𝑀
0
𝐶𝐹 1 = 0
𝐶𝐹 𝟐 = 0𝐶𝐹 𝝉 = 0
0
Events at component level, which do
not entail system failure No system
failure during first simulation
𝐶𝐹 1 = 0
𝐶𝐹 𝟐 = 0𝐶𝐹 𝝉 = 0
𝐶𝐹 𝜏 = 𝐶𝐹 𝜏 𝑓𝑜𝑟 ∀𝜏 ∈ [0, 𝑇𝑚]
𝑇𝑀
45
Example: System Reliability Estimation
• MC simulation of the first system life
• MC simulation of the second system life
0
𝐶𝐹 1 = 0
𝐶𝐹 𝟐 = 0𝐶𝐹 𝝉 = 0
𝐶𝐹 𝜏 = 𝐶𝐹 𝜏 + 1 𝑓𝑜𝑟 ∀𝜏 > 𝜏2
𝐶𝐹 1 = 0𝐶𝐹 𝟐 = 0 𝐶𝐹 𝝉𝟐 = 𝐶𝐹 𝝉𝟐 +1=1
Event at component level,
which do not entail system failure
Event at component level,
which causes the system failure
𝑇𝑀
𝑇𝑀
+1 +1 +1 +1
system failure at time 𝝉𝟐
𝝉𝟐
46
Example: System Reliability Estimation
• MC simulation of the first system life
• MC simulation of the second system life
• MC simulation of the M-th system life
0
𝐶𝐹 1 = 0
𝐶𝐹 𝟐 = 0𝐶𝐹 𝝉 = 0
𝑇𝑀
𝑇𝑀
+1 +1 +1 +1
𝑇𝑀
+1 +1
Event at component level,
which do not entail system failureEvent at component level,
which causes the system failure
𝐶𝐹 𝜏 = 𝐶𝐹 𝜏 + 1 𝑓𝑜𝑟 ∀𝜏 > 𝜏𝑀system failure at time 𝝉𝑴
𝝉𝑴
47
Example: System Reliability Estimation
• MC simulation of the first system life
• MC simulation of the second system life
• MC simulation of the M-th system life
• MC estimation of the system reliability
0
𝐶𝐹 1 = 0
𝐶𝐹 𝟐 = 0𝐶𝐹 𝝉 = 0
𝑇𝑀
𝑇𝑀
+1 +1 +1 +1
𝑇𝑀
+1 +1
Event at component level,
which do not entail system failureEvent at component level,
which causes the system failure
𝝉𝑴
𝐹𝑇 𝜏 ≅𝐶𝐹 𝜏
𝑀→ 𝑅 𝜏 = 1 −
𝐶𝐹 𝜏
𝑀
48
Example: System Reliability Estimation
t
F(t)=1-R(t)
4
1/4
1/2
49
SIMULATION FOR COMPONENT AVAILABILITY ESTIMATION
50
𝑞 =𝑀𝑇𝑇𝑅
𝑀𝑇𝑇𝑅 +𝑀𝑇𝑇𝐹
01
l
m
State X=0 (WORKING)
State X=1 (FAILED)
One component with exponential
distribution of the failure time
values
l 3· 10-3 h-1
m 25· 10-3 h-1
Limit unavailability:
?
51
01
l
m
One component with exponential
distribution of the failure time
values
l 3· 10-3 h-1
m 25· 10-3 h-1
Limit unavailability:State X=0 (WORKING)
State X=1 (FAILED)
𝑞 =𝑀𝑇𝑇𝑅
𝑀𝑇𝑇𝑅 +𝑀𝑇𝑇𝐹=
1𝜇
1𝜇+1𝜆
= 0.1071
52
01
l
m
State 1
State 0
One component with exponential
distribution of the failure and repair time
State X=0 (WORKING)
State X=1 (FAILED)
𝑿(𝒕)
53
01
l
m
State 1
State 0
One component with exponential distribution
of the failure and repair time
State X=0 (WORKING)
State X=1 (FAILED)
𝑿(𝒕)
54
01
l
m
State 1
State 0
One component with exponential
distribution of the failure and repair time
State X=0 (WORKING)
State X=1 (FAILED)
55Monte carlo simulation for estimating the
system availability at time t
• Divide the mission time, 𝑇𝑀, in bins and associate a
counter (of the system failure) to each bin:
𝐶𝑞 1 , 𝐶𝑞 2 ,… , 𝐶𝑞 𝜏 ,…• Initialize each counter to 0:
𝐶𝑞 1 = 0, 𝐶𝑞 2 ,= 0,… , 𝐶𝑞 𝜏 = 0,…
• If the component is failed in (𝑡𝑗 , 𝑡𝑗 + ∆𝑡) , thecorrresponding counter, cq(tj), increases cq(tj) = cq(tj) +1;otherwise, cq(tj) = cq(tj)
Trial 1
56Monte carlo simulation for estimating the
system availability at time t
• … another trial
Trial 1Trial 1
Trial i-th
57Monte carlo simulation for estimating the
system availability at time t
• … another trial
Trial 1
Trial i-th
Trial 1
Trial i-th
Trial M-th
58Monte carlo simulation for estimating the
system availability at time t
Trial 1
Trial i-th
Trial M-th
Trial 1
Trial i-th
Trial M-th
The counter cq(tj) adds 1 until M trials have been sampled
𝑞 𝑡𝑗 = 𝑃 𝑋 𝑡𝑗 = 1 ≅𝐶𝑞 𝑡𝑗
𝑀
q 𝑡𝑗 ≅𝐶𝑞 𝑡𝑗
𝑀
59
Pseudo Code
%Initialize parameters
Tm=mission time;
M=number of trials;
Dt=bin length;
Time_axis=0:Dt:Tm;
counter_q=zeros(length(1,Time_axis))
for i=1:M
%parameter initialization for each trial
t=0;
state=0; %state=working
while t<Tm
if state=0 %system is working sampling of failure time
t=t-[log(1-rand)]/lambda %failure time
state=1; % state=failed
failure_time=t;
lower_b=minimum(find(time_axis>=failure_time)); % first unavailability counter to be increased
else %system is failed sampling of repair time
t=t-[log(1-rand)]/mu;
state=1;
repair_time=t;
upper_b=minimum(find(time_axis > repair time)); %last unavailability counter to be increased
counter(lower_b:upper_b)= counter(lower_b:upper_b)+1; %increase all unavailability counter between lower_b and upper_b
end
end
end
Unav(:)=counter(:)/M
0 100 200 300 400 500 600 700 800 900 1000
0
0.2
0.4
0.6
0.8
1
Tiempo
Falta d
e d
isponib
ilidad
único ensayo
Simulacion Monte Carlo
Falta de disponibilidad en estado estacionario
Single trial
Monte Carlo simulation
Limit unavailability
Time
Una
va
ilab
ility
60
SIMULATION FOR SYSTEM AVAILABILITY / RELIABILITY
ESTIMATION
61
• T(tt’; k’)dt = conditional probability of a transition at tdt, given that thepreceding transition occurred at t’ and that the state thereby enteredwas k’.
• C(k k’; t) = conditional probability that the plant enters state k, giventhat a transition occurred at time t when the system was in state k’. Boththese probabilities form the ”trasport kernel”:
K(t; k t’; k’)dt = T(t t’; k’)dt C(k k’; t)
Stochastic Transitions: Governing
Probabilities
62
Phase Space
63
A
B
C
Components’ times of transition between states are exponentially distributed
( = rate of transition of component i going from its state ji to the state mi)i
mji 1l
Arrival
Init
ial
1 2 3
1 -
2 -
3 -
l)(
21BA
l)(
31BA
( )2 1A B
l l
)(32
BA
l)(
13BA
l)(
23BA
Indirect Monte Carlo: Example (1)
64
▪ The components are initially (t=0) in their nominal states (1,1,1)
▪ The failure configuration are (C in state 4:(*,*,4)) and (A and B in 3:
(3,3,*)).
Arrival
1 2 3 4
Init
ial
1 -
2 -
3 -
4 -
lC
21 lC
31 lC
41
lC
12 lC
32 lC
42
lC
13 lC
23 lC
43
lC
14 lC
24 lC
34
Indirect Monte Carlo: Example (2)
65
How to build 𝑇(𝑡|𝑡′ = 0, 𝑘′ = (1,1,1)?
The rate of transition of component A(B) out of its nominal state 1 is:
• The rate of transition of component C out of its nominal state 1 is:
•The rate of transition of the system out of its current configuration (1, 1, 1) is:
•We are now in the position of sampling the first system transition time t1, byapplying the inverse transform method:
where Rt ~ U[0,1)
lll)(
31)(
21)(
1BABABA
llllCCCC
4131211
llll CBA
1111,1,1
)1ln(1
1,1,101 Rtt tl
Monte Carlo Trial:
Sampling the time of transition
𝑇(𝑡|𝑡′ = 0, 𝑘′ = 1,1,1 = 𝜆 1,1,1 𝑒−𝜆1,1,1 𝑡
66
How to build 𝑇(𝑘|𝑡 = 𝑡1, 𝑘′ = (1,1,1))?
• Assuming that t1 < TM (otherwise we would proceed to the successive trial), wenow need to determine which transition has occurred, i.e. which component hasundergone the transition and to which arrival state.
• The probabilities of components A, B, C undergoing a transition out of theirinitial nominal states 1, given that a transition occurs at time t1, are:
• Thus, we can apply the inverse transform method to the discrete distribution
1,1,1
1
1,1,1
1
1,1,1
1 , ,l
l
l
l
l
lCBA
0
R
C 1,1,1
1
l
lA
1,1,1
1
l
lB
1,1,1
1
l
lC
C
10
Monte Carlo Trial:
Sampling the kind of Transition
𝑷(𝑨 ) 𝑷(𝑩) 𝑷(𝑪 )
= = =
67
• Given that at t1 component B undergoes a transition, its arrival state
can be sampled by applying the inverse transform method to the set of
discrete probabilities
of the mutually exclusive and exhaustive arrival states
l
l
l
lB
B
B
B
1
31
1
21 ,
0 1
R
S l
lB
B
1
31
l
lB
B
1
21
S
• As a result of this first transition, at t1 the system is operating in
configuration (1,3,1).
• The simulation now proceeds to sampling the next transition time t2 with
the updated transition rate
llll CBA
1311,3,1
Sampling the Kind of Transition (2)
RSU[0,1)
68
• The next transition, then, occurs at
where Rt ~ U[0,1).
•Assuming again that t2 < TM, the component undergoing the transition and
its final state are sampled as before by application of the inverse trasform
method to the appropriate discrete probabilities.
•The trial simulation then proceeds through the various transitions from one
system configuration to another up to the mission time TM.
)1ln(1
1,3,112 Rtt tl
Sampling the Next Transition
69
• When the system enters a failed configuration (*,*,4) or
(3,3,*), where the * denotes any state of the component,
tallies are appropriately collected for the unreliability and
instantaneous unavailability estimates (at discrete times tj [0, TM]);
• After performing a large number of trials M, we can
obtain estimates of the
• system unreliability at any time 𝑡𝑗 by
𝐹 𝑡𝑗 =𝐶𝐹(𝑡𝑗)
𝑀
• system instantaneous unavailability at any time 𝑡𝑗 by
q 𝑡𝑗 =𝐶𝑞(𝑡𝑗)
𝑀
Unreliability and Unavailability Estimation
70
Example: System Reliability Estimation
• Divide the mission time, 𝑇𝑀, in bins and associate a counter (of the system failure) to each bin:
𝐶𝐹 1 , 𝐶𝐹 2 ,… , 𝐶𝐹 𝜏 , …
• Initialize each counter to 0:
𝐶𝐹 1 = 0, 𝐶𝐹 2 ,= 0,… , 𝐶𝐹 𝜏 = 0,…
𝑇𝑀0
𝐶𝐹 1 = 0
𝐶𝐹 𝟐 = 0𝐶𝐹 𝝉 = 0
bin
71
( ) ( ) 0,R R
MC t C t t T
( ) ( ) 1 ,R R
MC t C t t T
( ) ( ) 1 ,R R
MC t C t t T
Example: System Reliability Estimation
Events at components
level, which do not entail
system failure
𝝉𝒓𝟏
𝝉𝒓𝟐
Repair Time
𝐶𝐹 𝜏 = 𝐶𝐹 𝜏 ∀𝜏𝜖[0, 𝑇𝑀]
𝐶𝐹 𝜏 = 𝐶𝐹 𝜏 + 1 ∀𝜏𝜖[𝜏1, 𝑇𝑀]
𝝉𝟏
𝝉𝟐
𝐶𝐹 𝜏 = 𝐶𝐹 𝜏 + 1 ∀𝜏𝜖[𝜏2, 𝑇𝑀]
𝐶𝐹 𝜏 = 𝐶𝐹 𝜏 ∀𝜏𝜖[0, 𝑇𝑀]
𝐹𝑇 𝜏 ≅𝐶𝐹 𝜏
𝑀→ 𝑅 𝜏 = 1 −
𝐶𝐹 𝜏
𝑀
72
Example: System Availability Estimation
• Divide the mission time, 𝑇𝑀, in bins and associate a counter (of the system failure) to each bin:
𝐶𝑞 1 , 𝐶𝑞 2 ,… , 𝐶𝑞 𝜏 , …
• Initialize each counter to 0:
𝐶𝑞 1 = 0, 𝐶𝑞 2 ,= 0,… , 𝐶𝑞 𝜏 = 0,…
𝑇𝑀0
𝐶𝒒 1 = 0
𝐶𝒒 𝟐 = 0𝐶𝒒 𝝉 = 0
bin
73
( ) ( ) 0,R R
MC t C t t T
Example: System Availability Estimation
Events at components
level, which do not entail
system failure
𝝉𝒓𝟏
𝝉𝒓𝟐
Repair Time
𝐶𝑞 𝜏 = 𝐶𝑞 𝜏 ∀𝜏𝜖[0, 𝑇𝑀]
𝐶𝑞 𝜏 = 𝐶𝑞 𝜏 + 1 ∀𝜏𝜖[𝜏1, 𝜏𝑟1]
𝝉𝟏
𝝉𝟐
𝐶𝑞 𝜏 = 𝐶𝑞 𝜏 ∀𝜏𝜖[0, 𝑇𝑀]
q 𝜏 ≅𝐶𝑞 𝜏
𝑀→ 𝑝 𝜏 = 1 −
𝐶𝑞 𝜏
𝑀
𝐶𝑞 𝜏 = 𝐶𝑞 𝜏 + 1 ∀𝜏𝜖[𝜏1, 𝜏𝑟2]
A real example of Monte Carlo Simulation
PRODUCTION AVAILABILITY EVALUATION OF AN
OFFSHORE INSTALLATIONZio, E., Baraldi, P., Patelli E. Assessment of the availability of an offshore installation by Monte Carlo simulation. International
Journal of Pressure Vessels and Piping 83 (2006) 312–320
75
2.2 MSm3/d
2.2 MSm3/d
4.4 MSm3/d
Production Gas : 5 MSm
3/d
Oil : 26500 Sm3/ d
Water : 8000 Sm3/d
4.4 MSm3/d
23300 Sm3/d
7000 Sm3/d
TC
TEG
Fuel Gas 25 b
50%
50%
Gas Lift 100b
Wells Separation
50%
50%
Gas Lift 60b
Gas Export 3.0 MSm
3/d, 60b
Oil export
TC
Flare
EC
Oil Trt °
TG
TG
Water Inj . Sea Wat . Trt °
13 MW
13 MW
7 MW
7 MW
6 MW
6 MW
0.1
MSm3
0.1
MSm3
0.1
MSm3
0.1
MSm3
1
MSm3/d
The offshore production plant
76
MONTE CARLO APPROACH
Associate a production level to
each of the 453 plant states A systematic
procedure
Production levels
…
oil gas water
Plant state ?TG1 TG2 TC1 TC2 TEG EC
77
A systematic procedure
7 different production
levels
Plant state
(Production
Level)
Gas
[kSm3/d]
Oil [k
m3/d]
Water
[m3/d]
0=(100%) 3000 23.3 7000
1 900 23.3 7000
2 2700 21.2 0
3 1000 21.2 0
4 2600 21.2 6400
5 900 21.2 6400
6 0 0 0Failure of TEG,
Failure of both TCs
Failure of both TGs
Failure of ony 1 TG
Failure of the EC
78Real system with preventive
maintenance
Production
level
Average
availability
0 8.13E-1
1 5.68E-2
2 6.58E-2
3 1.19E-2
4 3.55E-2
5 2.34E-3
6 1.50E-2
79
Mean Std
Oil [k m3/d] 22.60 0.42
Gas[k Sm3/d]
2687194.
3
Water[k m3/d]
6.04 0.76
Real system with preventive
maintenance
OIL
GAS
WATER
80Case A: perfect system and preventive
maintenances
Type of maintenance Frequency [hours] Duration [hours]
Type 1 2160 (90 days) 4
Type 2 8760 (1 year) 120 (5 days) Turbo-Generator and
Turbo-Compressors Type 3 43800 (5 years) 672 (4 weeks)
Electro Compressor Type 4 2666 113
81
Mean Std
Oil [k
m3/d]23.230 0.263
Gas[k Sm3/d]
2929 194.0
Water[k m3/d]
6.811 0.883
P.Maintenance Type 1 (TC,TG)
P.Maintenance Type 2 (EC)
P.Maintenance Type 3(TC,TG)
Case A: perfect system and preventive
maintenances
OIL
GAS
WATER
82Case B: Component Failure and no
preventive maintenances
83
Conclusions
• Complex multi-state system with maintenance and operational loops
MC simulation
• Systematic procedure to assign a production level to each configuration
• Investigation of effects maintenance on production
oil gas water
84
Where to study?
• Slides
• Book:
• Chapter 1
• Sections 3.1,3.3.1,3.3.2 (no examples of the multivariate normal distribution)
• Sections 4.1,4.2,4.3,4.4,4.4.1
• Section 5.1