280409 ISARET Presentation Slides
-
Upload
rafidah-rashid -
Category
Documents
-
view
226 -
download
0
Transcript of 280409 ISARET Presentation Slides
-
8/6/2019 280409 ISARET Presentation Slides
1/45
ISARET 2009
April 28th - 29th 2009
Data Monitoring and
Reconciliation for RefineryHydrogen Networks
by
Siti Rafidah Ab. RashidFaculty of Chemical Engineering
Universiti Teknologi MARA
-
8/6/2019 280409 ISARET Presentation Slides
2/45
ISARET 2009
April 28th -29th 2009
Definition of Terminologies
Data Reconciliation
an adjusting process of data to satisfy
process model constraints (material and
energy balances)
it estimates process variables by adjusting the
measurements and improves the accuracy of
the process variables
-
8/6/2019 280409 ISARET Presentation Slides
3/45
ISARET 2009
April 28th -29th 2009
Definition of Terminologies
Systematic Errors
the undetected mistakes that cause a
measurement to be very much farther from
mean measurement
caused by fouling of the sensors, wear and
tear, solid deposition on the probe, corrosion
on the sensors, miscalibration and instrumentmalfunction
-
8/6/2019 280409 ISARET Presentation Slides
4/45
ISARET 2009
April 28th -29th 2009
Definition of Terminologies
Refinery Hydrogen Network
the distribution path of the supply and demand of
hydrogen in a refinery
F-304
F-304 D-304
D-304
D-407
D-407 D-404
D-404D-409
D-409D-408
D-408
C-301A
C-402
C-401
FT903
FT004
FT035
FT025
FT025
D-305
D-305
PV361
PV361
FT011
FT028
FT028
FT049
FT
014
FT
014
N2Not Compensated
Not Compensated
Not Compensated
Not CompensatedNot CompensatedT & P
Compensated
Not Compensated
D-410
D-410 D-418
D-418
38oC 38oC 38oC
From E -408
To H -403
Vent
Flare
To D -2102
To D -302
From D -302
To D -408
38oC
73oC
38oC 38oC 44oC
Chloride
absorbedFrom
U1400 To
D-808
From
E-185
T & P
Compensated
38oC
73oC
38oC
44oC
38oC 38oC
70oC
38oC40oC
FI001
Vent
FromHydrogenHeader
-
8/6/2019 280409 ISARET Presentation Slides
5/45
ISARET 2009
April 28th -29th 2009
Background
The ISSUES are
Imbalance & Inconsistent Overall Material Balance
- Measuredstreams are corrupted with errors duringmeasurement, processing and transmission of themeasured signals
- Some streams are not been measuredbecause high
accuracy flowmeters are usually expensive
-
8/6/2019 280409 ISARET Presentation Slides
6/45
ISARET 2009
April 28th -29th 2009
Background (cont.)
FT
FT
FT
FTFT
FT
FT
Contains Error
Contains Error
Unmeasured
Unmeasured
HYDROGEN
NETWORK
Imbalance &
InconsistentOverall
Material
Balance
-
8/6/2019 280409 ISARET Presentation Slides
7/45
ISARET 2009
April 28th -29th 2009
HOW to Overcome?
Data monitoring and reconciliation
(cheaper & simpler)
improve the accuracy of measurements
improve plants profitability and flexibility
Background (cont.)
-
8/6/2019 280409 ISARET Presentation Slides
8/45
ISARET 2009
April 28th -29th 2009
Problem Statement
Strict environmental regulations & heavier crude oilsupply increases hydrogen demand
Hydrogen Network Management identifies the best routeto an optimised hydrogen network
However, not all process data are reliable or measured Therefore, there is a need to perform Data Monitoringand Reconciliation which will help the refiners toexecute hydrogen network mgmt effectively
In this work
A linear data reconciliation program is developed
&
an existing oil refinery hydrogen network is selected as a case study
-
8/6/2019 280409 ISARET Presentation Slides
9/45
ISARET 2009
April 28th -29th 2009
Objectives of the Study
The objectives of this project are to:
develop a systematic approach of data
reconciliation
apply the developed technique to a case
study of a refinery hydrogen networks
-
8/6/2019 280409 ISARET Presentation Slides
10/45
ISARET 2009
April 28th -29th 2009
Scope of the Study
This work only considers:
reconciliation of data for steady state
process single variable or linear system
(flowrate)
material balance as constraint
-
8/6/2019 280409 ISARET Presentation Slides
11/45
ISARET 2009
April 28th -29th 2009
Previous Works
Sanchez & Romagnoli (1996), utilized the QRfactorization to solve linear & bilinear datareconciliation problems. It allows the problem todecompose into lower dimension sub-problems.
Ripps (1965) continued the work of Reilly and
Carpani (1963) in systematic error detection.They defined a statistical method namely asGlobal Test.
-
8/6/2019 280409 ISARET Presentation Slides
12/45
ISARET 2009
April 28th -29th 2009
Previous Works (cont.)
Tariq (2006) has done a linear and
bilinear steady state data reconciliation on
a refinery hydrogen network using QR
Decomposition technique. He usedMATLAB as the calculations tool.
-
8/6/2019 280409 ISARET Presentation Slides
13/45
ISARET 2009
April 28th -29th 2009
Methodology
S4: Construction
of matrix Ax & Au
S1: Steady state,linear system
S3: Measured vs
Unmeasured etc.
S7: Method by Ripps
(1965)
S5: Method by
Sanchez & Romagnoli
(1996)
S2: Flowrates and
variance are data
available
S6: Method by
Narasimhan &
Jordache (2000)
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
14/45
ISARET 2009
April 28th -29th 2009
The Case Study:
Refinery Hydrogen Network
A refinery hydrogen network encountered a problem ofimbalance or inconsistent in material balance around thesystem. The study was initiated in 2004 and datareconciliation was advised.
This is crucial for future projects such as future plantdebottlenecking, hydrogen recovery and optimisation inits consumption and production.
The error-free flows are desired to give true picture ofthe network.
This particular hydrogen network is just like any othermodern refineries hydrogen networks. Is has 2production units and 7 units that consume hydrogen.
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
15/45
ISARET 2009
April 28th -29th 2009
The Case Study:
Refinery Hydrogen Network
In this study, only one producer and oneconsumer are considered.
The producer chosen for this exercise isnaphtha catalytic reformer denoted as Unit 400.
The hydrogen produced from this unit has thepurity of 85 to 88 mole%. This gas goes to thehydrogen header and then distributed to itsconsumers.
The selected consumer in this case is dieseldesulphurisation unit, which consumes hydrogento remove sulphur compound impurities.
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
16/45
ISARET 2009
April 28th -29th 2009
Refinery Hydrogen Networks
F-304F-304 D-304D-304
D-404D-404
C-301A
C-401
FT025FT025
D-305D-305
PV361PV361
FT011FT011
FT028FT028
FT049
FT
014
FT
014
Not Compensated
Not CompensatedNot Compensated
T & P
Compensated
Not Compensated
D-410D-410 D-418D-418
From E-408
To H-403
To D-2102
To D-302
From D-302
To D-408
73oC
38oC 38oC 44oC
Absorbed FromU1400 To
D-808
From
E-185
T & P
Compensated
73oC
38oC
44oC
38oC 38oC
70oC
38oC 4
0oC
From Storage
FI001
FT035
FT004
Not Compensated
FT903
Not Compensated
From
Hydrogen
Header
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
17/45
ISARET 2009
April 28th -29th 2009
Methodology
S4: Construction
of matrix Ax & Au
S1: Steady state,linear system
S3: Measured vs
Unmeasured etc.
S7: Method by Ripps
(1965)
S5: Method by
Sanchez & Romagnoli
(1996)
S2: Flowrates and
variance are data
available
S6: Method by
Narasimhan &
Jordache (2000)
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
18/45
ISARET 2009
April 28th -29th 2009
Step 1: Identify & Define ProblemSimplified Block Diagram of H2 Networks
M01
M04
M28M07C06S05C27 M10 C11 S13 M14
M12
P02
C03
S08 S09
S25M24C23M22S21
P26
S01
S11
S08
S07
S05S04
S02
S10
S09
P29
S24
S23
S26
S25
S28
S50
S49
S48S47S45S43S41
S39
S44
S42
S21S18S53
S15
S12 S13 S52 S16 S17 S19 S20
S22
S27
S51
S46
M01
M04
M28M07C06S05C27 M10 C11 S13 M14
M12
C03
S08 S09
S25M24C23M22S21
P26
S01
S11
S08
S07
S05S04
S02
S10
S09
P29
S24
S23
S26
S25
S28
S50
S49
S48S47S45S43S41
S39
S44
S42
S21S18S53
S15
S12 S13 S52 S16 S17 S19 S20
S27
S51
S46
FT
028
FT
014
FT
049
FT
035
FT
004 FT
903
FT
011
FT
025
FT
001
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
19/45
ISARET 2009
April 28th -29th 2009Step 2 & 3 : Analyse Available Plant
Data and Classify VariablesMEASURED VARIANCE ASSIGNED
VALUES VALUES
1.219 0.000016729 027.63 0.009089739 0
1.219 0.000016729 0
1.219 0.000016729 0.00701
0.01 0.000354655
1.209 0.000354655
26.412 0.009210576
26.412 0.009210576
2.476 0.0006212972.228 0.000621297
2.228 0.000621297
0.2476 0.000621297
3.976 0.000621297
3.933 0.000354655
0.0437 0.000035466
2.724 0.000354655
28.55 0.00035465525.416 0.000354655
1.728 0.000160437
0.06998 0.000002093
1.658 0.000146515
15.332 0.007737293
0.317 0.000003675
16.91 0.010339947
1.895 0.000621297
2.476 0.000621297
1.457 0.000621297
y = Measured
Values
unr =assigned
values
var = variances
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
20/45
ISARET 2009
April 28th -29th 2009
Step 4 (a): Process Model as ConstraintsElements of Measured Variable Matrix,Ax
Ax = S01 S02 S04 S05 S07 S08 S09 S10 S11 S12 S13 S15 S16 S17 S18 S19 S20 S21 S22 S23 S26 S42 S48 S49 S50 S52 S53
M01 1 -1 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
P02 -1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C03 0 0 -1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
M04 0 0 0 -1 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
S05 0 0 0 0 0 0 0 0 1 -1 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C06 0 0 0 0 0 0 0 0 0 1 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
M07 0 0 0 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 0
S08 0 0 0 0 0 0 0 0 0 0 0 0 1 -1 -1 0 0 0 0 0 0 0 0 1 0 0 0
S09 0 0 0 0 0 -1 0 0 0 0 0 0 0 1 0 -1 0 0 0 0 0 0 0 0 0 0 0
M10 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 1 0 0 0 0 0 0 0 0 0
C11 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 1 0 -1 0 0 0 0 0 0 0 0
M12 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0
C13 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 -1 -1 0 0 0 0 0 0
M14 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0
S21 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0
M22 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
C23 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
M24 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
S25 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0
P26 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 -1 1 1 -1 0 0
C27 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
M28 0 0 0 0 0 0 0 0 0 0 0 0 -1 0 0 0 0 0 0 0 0 0 0 0 0 1 1
P29 0 0 0 0 0 0 -1 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
23 units and 27 measured streams
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
21/45
ISARET 2009
April 28th -29th 2009
Example of process model
construction
For example, for the second unit P02, the
element of matrixAxcan be written as:
Ax= [-1 0 1 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0]
Input flow: S04
Output flow: S01
The other 25 streams are not associated to P02.
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
22/45
ISARET 2009
April 28th -29th 2009
Step 4 (b): Process Model as ConstraintsElements of Unmeasured Variable Matrix,Au
Au = S24 S25 S27 S28 S39 S41 S43 S44 S45 S46 S37 S51
M01 0 0 0 0 0 0 0 0 0 0 0 0
P02 0 0 0 0 0 0 0 0 0 0 0 0
C03 0 0 0 0 0 0 0 0 0 0 0 0
M04 0 0 0 0 0 0 0 0 0 0 0 0
S05 0 0 0 0 0 0 0 0 0 0 0 0
C06 0 0 0 0 0 0 0 0 0 0 0 0
M07 0 0 0 0 0 0 0 0 0 0 0 0
S08 0 0 0 0 0 0 0 0 0 0 0 0
S09 0 0 0 0 0 0 0 0 0 0 0 0
M10 0 0 0 0 0 0 0 0 0 0 0 0
C11 0 0 0 0 0 0 0 0 0 0 0 0
M12 -1 1 0 0 0 0 0 0 0 0 0 0
C13 0 0 0 0 0 0 0 0 0 0 0 0
M14 0 0 -1 1 0 0 0 0 0 0 0 0
S21 0 0 0 0 1 -1 0 0 0 0 0 0
M22 0 0 0 0 0 1 -1 1 0 0 0 0
C23 0 0 0 0 0 0 0 0 -1 -1 0 0M24 0 0 0 0 0 0 0 0 1 0 -1 0
S25 0 0 0 -1 0 0 0 0 0 0 1 0
P26 0 0 0 0 0 0 0 0 0 0 0 0
C27 0 0 0 0 0 0 0 0 0 1 0 -1
M28 0 0 0 0 0 0 0 0 0 0 0 0
P29 0 0 0 0 0 0 0 0 0 0 0 0
12 X 27 (12 units and 27 measured streams).
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
23/45
ISARET 2009
April 28th -29th 2009
Step 5: QR Decomposition
A decomposition of general rectangular matrixA, defined as:
AP = QR
where
Q is an orthogonal matrix
Ris an upper-triangular Pis a permutation matrix. Permutation matrix is a matrix that has
exactly one entry 1 in each row and each column and 0's elsewhere
In this work, Householder transformation technique is chosen
Alternatives are Givens transformations and Gram-Schmidt
orthogonalisation method
It is chosen because it is stable and easy to code in Visual Basic
Programming (Gunter and Van De Geijn, 2001) & QR decomposition was
noted as the most computationally efficient (Kelly, 1999)
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
24/45
ISARET 2009
April 28th -29th 2009
Step 5: Linear Data Reconciliation
Method by Sanchez and Romagnoli
(1996) is applied.
Where,
is reconciled values
Pis permutation matrix
y is the measured flowrate
x
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
25/45
ISARET 2009
April 28th -29th 2009
Step 6: Estimate the Unmeasured
Flowrates
Method by Narasimhan and Jordache (2000) isapplied.
Where,
R1, R2are subsets of matrix R
Q1 is subset of matrix Q
is reconciled measured variable
unris assigned valuesx
x
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
26/45
ISARET 2009
April 28th -29th 2009
Step 7: Systematic Error Detection
Method by Ripps (1965) is applied namely as GlobalTest.
Where, ris the vector of balance residuals, which is given by
r= Ay c
Ais the linear constraint matrix,Ax
ccontains known coefficients and for linear flowprocesses, c = 0
Vis variance-covariance matrix
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
27/45
ISARET 2009
April 28th -29th 2009
Step 7: Systematic Error Detection
(cont.)
is equal to the sum square of the
differences between the reconciled and
measured values. (Narasimhan and
Jordache, 2000).
= ( - y)2x
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
28/45
ISARET 2009
April 28th -29th 2009
Step 7: Systematic Error Detection
(cont.)
Under Ho, the above statistic follows a distribution with degrees of freedom,
where is the rank of matrixA. If the test criterion chosen is (where it is the
critical value of chi-square distribution at the chosen level of significance) then
Ho is rejected and a systematic error is detected if . also be
written as critical value of global testing, c.
2
2
In this work, theDOF = 23
(no. of units in the
system)
v = 10%
(commonly used)
c = 32.01Systematic
error is
detected
as = 225.5
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
29/45
April 28th -29th 2009
Results and Discussions
There are some differences between the results obtained and Tariqs (2006) The average differences is 0.9%. This shows the VBA Excel program
developed is acceptable and resulting very small differences only.
Tariq (2006) used MATLAB as his calculations tool
Full QR Decomposition which produces an upper triangular matrix R of the samedimension as A, and a unitary matrix Q so that A=Q*R.
This work uses VBA Excel as the calculations tool
Economy-Size QR Decomposition is applied which it only computes the first ncolumns ofQ and R is n-by-n for m n matrix A.
Similar to Tariqs, this program detected the presence of systematic error butdid not locate the error, which requires other techniques, beyond the scope ofthis project, to improve the data.
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
30/45
April 28th -29th 2009
Results and Discussions (cont.)
Comparison of Linear Data Reconciliation Results
0
5
10
15
20
25
30
1 3 5 7 9 11 13 15 17 19 21 23 25 27
Stream No.
Flowrates(ton/hr
)
VBA Excel
MATLAB
Possibly contains systematic error asdetected by the program developed.
Due to the scope of this work, the
program is detecting the presence of
systematic error of the system but
unable to identify the causes of the
error.
ISARET 2009
R lt & Di i ( t )
-
8/6/2019 280409 ISARET Presentation Slides
31/45
April 28th -29th 2009Results & Discussions (cont.)
Stream No. StreamName
Measured Flow Rates(ton/hr)
Reconciled Flow Rates (ton/hr) Variance(ton/hr)
1 S01 1.2190 1.2190 0.00002 S02 27.6300 27.6300 0.0000
3 S04 1.2190 1.2206 -0.0016
4 S05 1.2190 1.2021 0.0169
5 S07 0.0100 0.03324 -0.0232
6 S08 1.2109 1.5314 -0.3205
7 S09 26.4120 26.4120 0.0000
8 S10 26.4120 26.4120 0.0000
9 S11 2.4760 2.7048 -0.2288
10 S12 2.2280 2.4570 -0.2290
11 S13 2.2280 1.9992 0.2288
12 S15 0.2476 0.2478 -0.0002
13 S16 3.9760 3.0867 0.8893
14 S17 3.9330 4.4408 -0.5078
15 S18 0.0437 0.0945 -0.0508
16 S19 2.7240 2.7240 0.0000
17 S20 28.5500 26.9830 1.5670
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
32/45
April 28th -29th 2009
Results & Discussions (cont.)
18 S21 25.4160 26.9830 -1.5670
19 S22 1.7280 1.7280 0.0000
20 S23 0.0700 0.0700 0.0000
21 S26 1.6580 1.6580 0.0000
22 S42 15.3320 15.3320 0.0000
23 S48 0.3170 0.3170 0.0000
24 S49 16.9100 2.1060 14.804
25 S50 1.8950 1.8950 0.0000
26 S52 2.4760 2.2470 0.2290
27 S53 1.4570 1.4570 0.0000
Possibly contains systematic error
as detected by the program
developed. Due to the scope of
this work, the program is detecting
the presence of systematic error of
the system but unable to identify
the location of the error.
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
33/45
April 28th -29th 2009
Conclusions
As per objective of this study, a systematicapproach of linear data reconciliation isdeveloped.
Linear data reconciliation was done in the VBA
Excel program using QR decomposition method. The program is able to reconcile measured
variables, estimate unmeasured variables anddetect the presence of gross error.
The results of the program were compared withMATLAB. This is essential to ensure themathematics is correct through out the work.
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
34/45
April 28th -29th 2009
Future Works
This work can be extended by: Developing data reconciliation for bilinear and nonlinear
systems. Nonlinear system involves more than one variable, for instance,
combinations of two or more variables ie. flowrate, composition and
temperature. Energy and component balance can be considered for non-linear
system as constraints as addition to material balance.
Establishment of a program to identify the location of gross erroris also essential. The information can be used by engineers forscheduling instrumentation calibration and/or maintenance.
Current trend of data reconciliation also focuses on dynamicreconciliation. This can also be considered for future work.
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
35/45
April 28th -29th 2009
End of Presentation
Thank You
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
36/45
April 28th -29th 2009
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
37/45
April 28th -29th 2009
Attachments
ISARET 2009
-
8/6/2019 280409 ISARET Presentation Slides
38/45
April 28th -29th 2009
Attachment: GUI of the program
developed
ISARET 2009
A il 28th 29th 2009
-
8/6/2019 280409 ISARET Presentation Slides
39/45
April 28th -29th 2009
Attachment: Algorithm of Data
Reconciliation Program
Start
Stop
Input is
unmeasured flow
matrix, Au
Solve QR decomposition of
unmeasured matrix, Au,
using equations (3.11)
through (3.14)
Output are Q and
R matrices of
decomposed Au
ISARET 2009
A il 28th 29th 2009
-
8/6/2019 280409 ISARET Presentation Slides
40/45
April 28th -29th 2009
Attachment:Algorithm of
Data
ReconciliationProgram
Start
Input are:
measured flowrates matrix, Ax
unmeasured flowrates matrix, Au
Q and R matricesvalues of measurements, Y
variance of measurements, var
assigned values, unr
Reconcile the flowrates using the equations
(3.18) to (3.23)
Output are:
Reconciled flowrates, x
Estimation of unmeasured flowrates, u
Global test parameter, ?
Systematic error
is detected
No systematic
error is detected
? > ? ?2
Stop
-
8/6/2019 280409 ISARET Presentation Slides
41/45
ISARET 2009
A il 28th 29th 2009
-
8/6/2019 280409 ISARET Presentation Slides
42/45
April 28th -29th 2009
Attachment: Variable Classifications
Redundant variable is a measured process variable that is over-determined if it can also be
computed from the balance equations and the rest of the measured variables.
Unmeasured variables can be grouped as observable or non-observable variables. Observable
variable is unmeasured variable that is determinable if it can be evaluated from the available
measurements using the balance equations.
ISARET 2009
April 28th 29th 2009Attachment:
-
8/6/2019 280409 ISARET Presentation Slides
43/45
April 28th -29th 2009Attachment:
Previous Works on QR Decomposition in
Data Reconciliation Swartz (1989) used the QR decomposition in the matrix projection toeliminate the unmeasured variables.
Madron (1992)proposed classifying measured & unmeasuredvariables of linear systems according to pre-established criteria ofrequired & nonrequired.
Sanchez & Romagnoli, (1996), utilized the QR factorization to solvelinear & bilinear data reconciliation problems. It allows the problem todecompose into lower dimension sub-problems.
Kelly (1998) used different matrix projection techniques & highlightedtwo simpler approaches to determine the matrix projection(introduced by Crowe et al. (1983)).
ISARET 2009
April 28th 29th 2009Att h t P i W k
-
8/6/2019 280409 ISARET Presentation Slides
44/45
April 28th -29th 2009Attachment: Previous Works on
Gross Error Detection
Reilly and Carpani (1963) formulated Global Test (detection test)which is the collective chi-square test of all the data and the univariate test for constraints, based on the normal distribution. Ripps (1965) proposed a method which eliminates the measurement that renders the largest
reduction in a test statistics until no test fails. Romagnoli and Sephanopolous (1981) developed a systematic strategy to locate the source of
gross error. The method also rectifies the gross and biased measurement errors in a chemicalprocess.
Almasy and Sztano (1975) presented a global test and measurement test that possessesmaximum power test when there is only one gross error in the measurements, and is called the
MPT. Mah et.al (1976) presentedThe Constraint and Nodal Test (detection and identification tests).
This method requires linear constraints and measured variables. Mah and Tamhane (1982) proposed the univariate measurement test, which examines each
measurement adjustment. The statistical test based on the adjustment distribution by which firstprocess data are reconciled.
Narasimhan and Mah (1987)proposed a test named Generalised Likelihood Ratios. GLR hasthe capability to identify the location of the error and differentiate the types of the error such asinstrument related error or process model related error.
Rollins and Davis(1992) introduced UBET (Unbiased Estimation Technique). UBET is limited tonormally distributed errors, steady state and linear constraints.
Tong and Crowe (1996) introduced Principal Component Analysis (PCA). This technique is a setof correlated variables is transform into a new set of uncorrelated variables. PCA is a veryeffective method for multivariate data analysis.
-
8/6/2019 280409 ISARET Presentation Slides
45/45