Download as pdf or txt
Download as pdf or txt
You are on page 1of 20

WATER RESOURCES SYSTEMS

ENGINEERING
WEEK13

Doosun Kang, Kyung Hee University


Uncertainty Analysis

2
Definition of Uncertainty
3

▪ Lack of perfect information regarding the phenomena, processes, and data


▪ The occurrence of events that are beyond one’s control

▪ 확실성의 결여 (사전적 의미)

▪ 시스템이 감당할 수 있는 범위를 벗어난 사건(event)의 발생가능성 (경영, 재난, 재해분야)

▪ 어떤 양에 대해서 관측하거나 계산한 값이 참값을 벗어나 있는 상태 (데이터 관측)

▪ 어떤 일이 벌어질 것인지 알지 못하는 상태 (실험 심리학)

▪ 대부분의 공학적인 설계, 운영, 의사결정 과정은 불확실성 하에서 이루어지는 경우가 대부분

▪ 그러므로, 시스템의 안정성, 신뢰도는 불확실성의 정도와 밀접한 관련


Sources of Uncertainties
4
Purpose of Uncertainty Analysis
5

▪ Model provides single best estimate of output variable assuming


input parameters are fixed and certain

▪ In practice, model input parameters are uncertain

▪ To quantify uncertainty of model outputs in terms of interval


(bounds, range) rather than single estimate

불확실성 분석의 목적은 시스템의 입력값 혹은 내부 변수들의 변화에 따라 발생가능한


시스템 결과값의 변동량을 정량화하기 위함

Rational formula: Q = 0.2778CIA


D-W equation: HL = f*(L/D)*(V2/2g)
Uncertain input parameters in hydrosystems
6

Watershed System Pipe System


Output – Runoff Output - Pressure
Rainfall Diameter (D)
Land use & Soil type Roughness coef. (C)
Basin area Consumer’s demand (q)
Uncertainty Analysis in Pipe Network
7
D-W equation: HL = f*(L/D)*(V2/2g)

59

System Input & Parameter:


65

- Diameter [D-σ ~ D]

26
- Roughness coef. [f-σ ~ f+σ]
- Node demand [q-σ ~ q+σ]
23
Booster 1

44

Tank 1 Booster 2

Legend

Valve

System Output:
62 Tank 2
Booster

Booster 3
- Pipe flow [Q-σ ~ Q+σ]
Pump
Treatment Plant

56 - Node pressure [P-σ ~ P+σ]


- Water quality [WQ-σ ~ WQ+σ]
Uncertainty Analysis in Pipe Network
8 Uncertainty Level 1 Unce

0.12 0.32

Probability density
0.24
59 0.08

<< 0.16
Low
0.04
65 0.08

0 0
65 70 75 80 85 90 95 10 20

Uncertainty Level 2 Unce


26
0.08 0.2
Min. required pres.
23
Booster 1

Probability density
0.16
0.06
(최소요구수압)
<< 0.12
Medium
44

Tank 1
0.04
Booster 2 0.08
Legend 0.02
0.04
Valve 0 0
62 Tank 2 65 70 75 80 85 90 95 10 20
Booster

Booster 3 Uncertainty Level 3 Unce


Pump
Treatment Plant 0.05 0.16

Probability density
0.04
0.12
56 0.03
<<0.08
High
0.02
0.04
0.01

0 0
65 70 75 80 85 90 95 10 20
Pressure (psi) Wat
MCS histogram FOSM
Uncertainty Analysis in Pipe Network
9 120
Uncertainty Level 1
60
Unc

100 50

Water age (hrs)


Pressure (psi)
80 40

59 60 30

40 << Low
20
65 20 10

0 0
12am 4am 8am 12pm 4pm 8pm 12am 12am 4am 8am

Uncertainty Level 2 Unc


26 120 60
23
Booster 1 100 50

Water age (hrs)


Pressure (psi)
<< Medium
44 80 40

60 30
Tank 1 Booster 2
40 20
Legend
Min. required pres.
20 10
Valve
62 Tank 2 0 0
Booster 12am 4am 8am 12pm 4pm (최소요구수압)
8pm12am 12am 4am 8am

Booster 3 Pump Uncertainty Level 3 Unc


Treatment Plant
120 60

100 50

Water age (hrs)


Pressure (psi)
56 80
<< High
40

60 30

40 20

20 10

0 0
12am 4am 8am 12pm 4pm 8pm 12am 12am 4am 8am
System Design considering Uncertainty
10

 Reliable(robust) system – A system completing its expected functions


against internal/external disturbance (uncertainty)

 내구성이 강한 시스템 – 외부변화요인(불확실성)에도 시스템 본연의 기능을 실패 없이 수행하는


시스템

System Reliability
Probability Density

System Reliability (0~1)

Min. Required
Pres.
Pressure
System Design considering Uncertainty
11

 Reliable(robust) system – A
59

system completing its expected


functions against uncertainty
65

26
23
Booster 1

44

Objectives of Design:
Tank 1 Booster 2

Legend

62 Tank 2
Valve
Booster
Obj. 1 – Min. Cost
Booster 3 Pump
Treatment Plant Obj. 2 – Max. Reliability

56
Monte Carlo simulation (MCS) uses
repeated random sampling to determine
the potential ranges of outputs

Monte Carlo Simulation (MCS)

Uncertainty quantification method


Most widely-used,
Most accurate,
BUT, most time consuming

12
Random number generation 1
13

Normal distribution
normrnd(mu,sigma) generates random numbers from the normal
distribution with mean parameter mu and standard deviation
parameter sigma

>> x1 = normrnd(50, 5, 5000, 1);


>> x1 contains random numbers following normal distribution with mean=50, stdv=5

>> Confirm using histogram


800

>> hist(x1,20)
700

600

500

400

300

200

100

0
30 35 40 45 50 55 60 65 70
Random number generation 2
14

Uniform distribution
unifrnd(A, B) returns an array random numbers generated from
the continuous uniform distributions with lower and upper endpoints
specified by A and B, respectively.

>> x2=unifrnd(40, 50, 5000, 1);


>> x2 contains random numbers following uniform distribution with range [40, 50]

>> Confirm using histogram


300

>> hist(x2,20) 250

200

150

100

50

0
40 41 42 43 44 45 46 47 48 49 50
Previously optimized pipe network
15 q=0.1cms q=0.1cms
6
3 5 6

3 5 q=0.1cms
Source
P 1 2 4 7
1 2 4 7

Final Solutions
Network Infos: Optimal Diameter:

- q = 0.1cms for nodes 3, 6, and 7 400-400-300-400-350-350-300 mm


Optimal Pumping Head:
- L = 1000m for all pipes
60m
- Source head = 100m
- Node elevation = 110m for all nodes Nodal pressure with no uncertainty
- Use Darcy-Weisbach equation P (node3) = 14.11m
- Friction factor, f = 0.02 for all pipes P (node6) = 8.16m

- Min. required pressure = 5m P (node7) = 7.66m


Uncertainty quantification using MCS
16

Uncertain Input Parameters:

1. Pipe diameter – uniform distribution


D = [D-∆D ~ D], ex) [0.95*D ~ 1.0*D]

2. Pipe roughness coefficient – normal distribution


f = mean ± k*stdv, ex) stdv = 0.05*mean

3. Nodal demand – normal distribution


q = mean ± k*stdv, ex) stdv = 0.05*mean
Uncertainty in nodal pressure
17

0.2 0.2 0.2

0.18 Node 3 0.18 Node 6 0.18 Node 7


0.16 0.16 0.16

0.14 0.14 0.14


5m
probability density

probability density

probability density
0.12 0.12 0.12

0.1 0.1 0.1

0.08 0.08 0.08

0.06 0.06 0.06

0.04 0.04 0.04

0.02 0.02 0.02

0 0 0
-30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30 -30 -20 -10 0 10 20 30
Pressure (m) Pressure (m) Pressure (m)

f(x)
3 5 6 6 Reliability(0~1)

Source 3 5
P 1 1 2 2 4 4 7 7 Pres
How to improve system reliability?
18

1. Increase pipe diameter


2. Increase pumping head

Trade-off
System cost vs. System reliability
(Objective 1) (Objective 2)
Homework #8 – Pipe Network Design w/ Uncertainty
19

I will give you incomplete codes.

Task1) You should complete the codes and make them run.

Task2) Improve system design to satisfy system reliability:


- Find least-cost design to achieve nodal reliability of 0.6 for all locations
- Find least-cost design to achieve nodal reliability of 0.7 for all locations
- Find least-cost design to achieve nodal reliability of 0.8 for all locations

Note) For Task2, you must use a GA


Tip) You can set a cost as an objective function and reliability as a constraint
Remaining classes
20

▪ Week 14 (12/4) – Homeworks #6, #7, #8 review

▪ Week 15 (12/11) – Final exam

You might also like