ISSN 0104-6632 Printed in Brazil
www.abeq.org.br/bjche
Vol. 32, No. 04, pp. 919 - 927, October - December, 2015 dx.doi.org/10.1590/0104-6632.20150324s00001038
*To whom correspondence should be addressed
Brazilian Journal
of Chemical
Engineering
PLANTWIDE PERIODICAL DISTURBANCES
ISOLATION AND ELIMINATION IN A
PETROCHEMICAL UNIT
M. Farenzena
1*, C. A. Kayser
2and J. O. Trierweiler
11Chemical Engineering Department, Federal University of Rio Grande do Sul,
(UFRGS), Rua Luiz Englert s/n, Campus Central, CEP: 90.040-040, Porto Alegre - RS, Brazil.
Phone: + (55) (51) 3308 3918 E-mail: farenz@enq.ufrgs.br; jorge@enq.ufrgs.br
2TriSolutions - Engineering Solutions, Rua Gal. Bento Martins, 24 cj 1101, Centro,
Porto Alegre, RS, CEP: 90010-080 Brazil. Phone: + (55) (51) 3227 8514 E-mail: cristine@trisolutions.com.br
(Submitted: March 3, 2011 ; Revised: January 23, 2015 ; Accepted: January 26, 2015)
Abstract - Reducing process variability is crucial to reach a more profitable operating point. Periodical disturbances, however, impose barriers to achieve this goal. Their effect can be strong since one disturbance that appears in a specific loop of a highly coupled plant can be seen in several loops. Thus, isolating their source and diagnosing their cause are essential. In this work, we describe the application of spectral independent component analysis to isolate a periodical disturbance that has a strong impact on the final variability in a polyethylene plant located in Southern Brazil. After the first analysis, the source was detected and the cause identified: valve stiction. To identify the cause (valve, bad tuning, or periodic disturbance), we used the methodology based on higher-order statistics. Once the valve problem had been overcome, the product variance was reduced by 93%.
Keywords: Fault diagnosis; Plant-wide disturbance; Oscillation; Independent Component Analysis.
INTRODUCTION
One frequent cause of poor process performance is the presence of plant-wide periodical disturbances (Thornhill and Horch, 2007) whose effect can spread through the entire plant, inhibiting the process from achieving a more profitable operating point. Nowa-days, plants are more coupled and have a large num-ber of recycles because of mass and heat integration. One oscillation that starts in a specific loop can be propagated to the entire process, increasing the product variability. Thus, it is clear that detecting and elimi-nating plant-wide oscillations are essential to ensure
process profitability and reduce product variability. However, the diagnostics of the loop that is the source of the disturbance and the cause of the oscillation is not straightforward.
This is the scenario seen in a petrochemical plant located in Southern Brazil. One periodical oscillation affects the product variability and, because of plant recycles, the engineers could not detect either the source or the cause. Our goal is to detect and find the source of the oscillation, eliminating it (if possible) after diagnosing the cause.
92 lo ti m en m T in an T (A sp w in th th fo p m et (J 2 H et co 20 1. Detectin 2. Detectin oop where th
3. Diagnos ion or poor tu Initially, th many cases,
ngineers. In methods to a The first is
ntegral of ab nd zero cros The second is ACF) (Thorn pectral peak wide periodic n Thornhill a
The secon he oscillation he literature ocus which ology-based methods are b t al., 2004; Jiang et al., 2 007). Good Hagglund (20 t al. (2014). omponent m
Once the s
ng the oscilla ng the root-c he oscillation sing the caus uning. he disturban the oscillati the literatur automatically time-domain bsolute erro ssings (Thorn s based on th nhill et al., 2 detection. A cal disturban and Horch (2 nd step is to n, i.e. the so , there are are segment (Duan et a based on a no Thornhill, 2007; Xia an
reviews can 007), Choudh
In this wor method (Xia e source of the
ation; cause of osci n started); and
se of oscillat
nce should be ion is detect re, there are
y detect loo n methods ( or (IAE) (Hä
nhill and Hä he Autocovar 2003). The th A good revie nce detection
007). o identify the ource of the
several met ted into data al., 2014). T on-linearity t
2005) or sp nd Howell, 2 n be found in hury et al. (2 rk, the spectr et al., 2005a) e oscillation
Figure 1:
M. Farenzena, C
Brazilian Jou
illation (i.e., d
tion: valve s
e diagnosed. ted visually three groups op oscillatio (e.g., based ägglund, 19 ägglund, 199
riance Funct hird is based ew about pla n can be fou
e root cause disturbance. thods with t a-based and The data-ba test (Choudh pectral analy 2005; Xia et
n Thornhill a 2008) and Du ral independ ) will be used is isolated,
: Petrochemi
C. A. Kayser and
urnal of Chemica
the tic-. In by s of ons. on 95) 97). tion d on ant-und e of . In this to-sed hury ysis al., and uan dent d. the cau ing sti the the hu Sin of (20 agn che qu olo app sio ces list
ical plant sch
J. O. Trierweiler
al Engineering
use should b g or external ction is eva e cause is tu ere are many ury et al., 200
nghal and Sa these techni 014).
The paper s nose and elim emical plant uently. The n
ogies used in plication are ons are drawn
P
The unit stu sses: prepolymer polymerizat monomer re monomer re Its process t with the ma
hematic repre
be diagnosed l disturbance aluated; if th ning or distu y methods t 04; Horch, 19 alsbury, 2005 iques are su
sums these th minate period , whose desc next section n this work, a e then discu
n.
PLANT DES
udied in this
rization; tion; ecovery; ecycle to the
schematic is ain loops is g
esentation.
d: valve, or p es. First the he result is n
urbances. In to assess sti 999; Rossi an
5; Yamashita ummarized in
hree sets of m dic disturban cription is pr summarizes and the main ussed. Finally
SCRIPTION
paper consis
reaction stag
s shown in F given in Tabl
poor loop tu hypothesis negative, the
the literatur ction (Chou nd Scali, 200 a, 2006). Mo n Brásio et a methods to d nces in a petr
rovided subs s the metho n results of th
y, the concl
N
sts of four pr
ge.
T D D L L L L ca m an co p re at li re re
Table 1: List
Loop Funct
DC01 Mud C
DC02 Mud C
FC01 Monom FC02 Monom FC03 Monom Reacto FC04 Top St FC05 Steam FC06 Reflux FC07 Recirc FC08 Make-LC01 Reacto LC02 Flash D LC03 Recyc LC04 Feed T PC01 Recyc PC02 Feed T PC03 De-ga PC04 Reacto
Monomer atalyst and c merization re
nd unreacted ontinuously olymerizatio eactors, filled
ted in series. Each react ine with inde eactor) and F eceive remot
Plantw
Brazilian
t of loops in
tion
Concentration in Concentration in mer Inlet Flow mer Inlet Flow mer Inlet Flow or
tream of Flash D m to Recycle Mo
x to Recycle Mo culation Flow (M
-up of Fresh Mo ors Surge Drum Drum Level cle Monomer Sc Tank Level cle Monomer Sc Tank Pressure assing Stream Pr
ors Surge Drum
and the cata co-catalysts, actor. The m d monomer i
sent to pol on process co
d with liquid
tor has an in ependent flo FC03 (second te setpoints fr
wide Periodical D
Journal of Chem
the petroch
n First Polymer n Second Polym
to First Polym to Prepolymeri to Second Poly
Drum onomer Scrubbe onomer Scrubb Minimum Flow onomer m Level crubber Level crubber Pressur ressure m Level alytic compl are fed into mixture form
in the prepol lymerization onsists of tw d monomer,
ndependent m ow controller d reactor). Th from concentr Fi Disturbances Isola mical Engineering hemical plan rization Reactor merization Reac merization React ization Reactor ymerization er Reboiler ber w to Pump)
re
lex, formed o the prepol ed by polym lymerization
reactors. T wo tubular lo that are ope
monomer fe rs, FC01 (fi hese flow loo ration of slur
igure 2: Proc
ation and Elimina
g Vol. 32, No. 04,
nt. r ctor tor by ly-mer n is The op er-eed rst ops rry (m tro re in tio ce co co ch le th m re ac th dr ce fr m se of m Th of by bl di ad sm op th se cess variable
ation in a Petroch
, pp. 919 - 927,
monomer + p ollers, DC0 eactor). It is n slurry conc on increase, entration setp Slurry from ontinuously ond stage of harge is cont evel of the su he polymeriz monomer. Th
eactor is disc cted monom he polymer.
rum flows to ess, such as d
Unreacted rom the de-ga mer scrubber, eparated from
f the scrubb monomer fee he tank level f fresh mono y PC02 that v The analys les, as shown isturbances c dvisable to moother ope perating poin he oscillation econds, calle e periodogram hemical Unit
October - Decem
polymer) ins 1 (first rea
important to centration res as it permit point.
m the first fed to the s f the polyme trolled by LC urge drum, w zation reacto he resulting
harged to a f mer is evapo Polymer di o the next sta de-gassing.
monomer, assing proces
where entra m recovered ber is conde ed tank, whi l is controlled omer, while vaporizes pa sis in the per n in Figure 2 can be foun
remove som eration and a nt. In this w ns with perio
d respectivel
ms.
mber, 2015
side the reac ctor) and D o note that l sults in a po ts one to inc
polymerizat second react erization, the C01. This loo which must ors are fille product from flash drum w
rated and se ischarged fr ages of the pr
from the fla ss, flows to a ained polyme monomer. T ensed and p
ich supplies d by manipu the pressure art of the fres riodogram of 2, indicates t d in the pla me of them
achieve a m work, we des ods 320 sec ly dist1 and d
9 ctor loop co DC02 (secon
low variabili ssible produ rease the co
tion reactor or. In this s e process di op controls th
guarantee th ed with liqu m the secon where the unr
eparated fro rom the fla roduction pr
ash drum an recycle mon er particles a The top strea umped into s the reactor
lating the flo e is controlle sh liquid. f several vari
that periodic ant. Thus, it to achieve more profitab
sire to remov conds and 42
dist2.
922 M. Farenzena, C. A. Kayser and J. O. Trierweiler
Brazilian Journal of Chemical Engineering
METHODOLOGY
This section describes the methodologies used to automatically detect the oscillation, its root cause (source), and the cause of oscillation.
Oscillation Detection
Initially, the oscillation was automatically de-tected using the integral of the square error, a method proposed by Hägglund (1995).
The idea behind the method is simple: based on time trend zero-crossing, the integral of the absolute error between each zero crossing is computed (IAEC). It is then compared with a threshold value (IAELIM). If IAEC > IAELIM, then the process has an oscillatory behavior.
The oscillation detection procedure can be sum-marized as follows:
1. Choose an acceptable oscillation amplitude (a); 2. Compute IAELIM as 2a/. is 2/Ti and Ti is the integral time of the controller.
3. Monitor IAEC. Restart it when the control error changes its signal.
4. If IAEC exceeds IAELIM then the oscillation has occurred.
Detecting the Root Cause (Source) of Oscillation
To detect the root cause of the oscillation, the methodology based on Spectral Independent Compo-nent Analysis (SICA) was used (Xia et al., 2005). Initially, the time-domain ICA will be described and then the methodology to detect the root cause of oscillations will be explained.
Consider that the plant has m sensors, whose ob-servations are xm.
1 2
T
m
X x x x (1)
They are linear combinations of n independent, non-Gaussian source outputs. Each column is called an Independent Component (IC).
1 2
T
n
S s s s (2)
The matrix of observations (X) can be written as a linear function of the matrix of ICs (S).
X AS (3)
where A is called the mixing matrix (m by n). Each sensor can be decomposed into linear combinations of ICs.
,1 1 ,1 1 ,1 , 1
i i i i n
x a s a s a s i m (4) The ICA problem involves the estimation of both A and S. The Fast ICA algorithm (Hyvärinen et al., 2001) was used in this work.
In the spectral ICA model, the rows of X are single-sided power spectra P(f) over a range of frequencies (f) of the same sensor. P(f) can be estimated using Discrete Fourier Transform (DFT) (Oppenheim et al., 1999). The main advantage of using power spectra instead of time series is that the first is blind to the time delays. Besides, SICA can isolate a single peak in each independent component, when multiple os-cillations are present in the plant.
The procedure to apply the methodology based on SICA is described below:
1. Compute the power spectra for all measured variables (X).
2. Decompose X into independent components, obtaining A and S (using FastICA).
3. Find the sign of the dominant peak for each IC, denoted by SNj, (j=1…n);
4. Adjust the A and S matrixes using the follow-ing relation:
1 2
1 2
.
Y
n
n
B A diag SN SN SN diag SN SN SN S
(5)
Then
. X BY
Here, the new term, based on matrix B, called the significance index is introduced. It provides the im-portance of the combined matrix elements. Values close to 1 represent a strong impact from an IC in a power spectrum signature. Smaller values of the significance index show a smaller impact of a given IC.
Then, each IC was adjusted to achieve a maxi-mum significance index equal to 1.
1. Find the maximum absolute value for each col-umn of B (j, j=1…n);
2. Scale the mixing matrix B and IC matrix Y;
1 1 1
1 2
1 2
.
n
n
A B diag
C diag Y
(6)
Then
Plantwide Periodical Disturbances Isolation and Elimination in a Petrochemical Unit 923
Brazilian Journal of Chemical Engineering Vol. 32, No. 04, pp. 919 - 927, October - December, 2015
Based on the new matrix A, the source of each in-dependent component, or plant-wide disturbance, can be identified. The loop for a given IC whose significance index is close to 1 is probably the root cause of that oscillation.
In the work of Xia et al. (2005), the dominance of each independent component over the plant is ana-lyzed, helping to discover which are the main plant disturbances. However, in our case, a visual analysis showed that the mentioned oscillation (see Fig. 2) appeared in several loops, causing a strong impact in the product variability.
Detecting the Cause of Oscillation
The last step is to diagnose the cause of the oscilla-tion. In this work, we will apply the methodology based on higher-order statistics, proposed by Choud-hury et al. (2004). To corroborate the results, the meth-odology to quantify stiction based on ellipsis interpola-tion will also be used (Choudhury et al., 2008).
The method proposed by Choudhury et al. (2004) claims that, if the process is locally linear and a non-linear behavior is present, then the valve is responsi-ble for this behavior. If the loop has an oscillatory behavior and the valve is working properly, the prob-lem can be poor loop performance or other external disturbance (e.g., a disturbance transferred from an-other loop with oscillatory behavior).
Initially the non-Gaussianity index (NGI) is com-puted, using the bicoherence (bic2) concept.
2 1 2 2
2 2
1 2 1 2
,
B f f bic
E X f X f E X f f
(7)
where X(f) is the discrete Fourier transform of any time series x(k) and B(f1,f2) is called the bispectrum
in the frequencies f1 and f2. The bispectrum is the
third order cumulant in the frequency domain. It is defined as:
1, 2
1 2 1 2
*
B f f E X f X f X f f (8) where * denotes the complex conjugate. One positive feature of bic2 is that it is bounded between 0 and 1.
Assuming that bic2 at each frequency is a chi-squared (2) distributed variable with 2 degrees of freedom, a modified test formulated by averaging the squared bicoherence over the triangle of the principal domain with better statistical properties will be used to verify signal Gaussianity. The test can be summa-rized as follows:
Null hypothesis: the signal is Gaussian,
Alternate hypothesis: the signal is non-Gaussian.
Under the null hypothesis, the test can be based on the following equation:
2 2
2 αχ
P KLbîc c α (9)
where cαχ2 the critical calculated from the central ²
distribution table for a significance of and 2L de-grees of freedom, K is the number of data segmenta-tion during the bicoherence computasegmenta-tion, L is the number of bifrequencies inside the principal domain
of the bispectrum, and bîc2
iL1bîc2.If the signal is Gaussian, the process is assumed to be linear. If the signal is non-Gaussian, the process nonlinearity should be tested. To evaluate the non-linearity, the constancy of the squared bicoherence should be evaluated. In this work, the maximum squared bicoherence can be compared with the aver-age squared bicoherence plus two standard devia-tions. At 95% confidence level, if the maximum bi-coherence (bîcmax2 ) is less than 2
2
( 2 ),
bîc
bîc the bicoherence curve is assumed to be constant. The index to evaluate the nonlinearity is defined as:
2
2 2
2
max bîc
NLI bîc bîc (10)
Where bîc2 is the standard deviation of the squared bicoherence and bîc2 is the average of the squared bicoherence. If NLI = 0, the process is lin-ear; otherwise the process is nonlinear. If the loop is nonlinear, then the valve “suffers from” stiction.
To corroborate the diagnostics provided by the method based on higher-order statistics, the method based on ellipsis interpolation (Choudhury et al., 2006) will also be used to diagnose and quantify valve stiction. If the process variable (pv) and control output (op) plot has an ellipsis pattern, as shown in Figure 3, then the stiction is confirmed. The apparent stiction can be quantified as the length of the hori-zontal ellipsis axis (sb).
RESULTS
92 O th th w an co T IA ce p tu D v M w w (v W 24 Figure 3: Oscillation D The oscill he IAE met he following
191 for D seconds); 20 for PC seconds);
The IAEC whole dataset
nd 2nd IAEC ontrollers, as
Table 2: Lar AEC) for DC
Loop DC01 DC02 PC01 PC02 Based on eed the limit rocess is str urbances. Diagnosing th Initially, th ariables had Moreover, w without comp was computed version R11, We collected
pv versus op
Detection
lation is dete thod previou
IAELIM: DC01 and
C01 and PC
C values hav t and the tw
C) are shown s well as thei
rgest IAEC
C01, DC02, P
1st IAE C
1900 1730 117 215
Table 2, we ting IAELIM v rongly affect
he Oscillatio
he data were the same sa e ensured th pression. Th d using the p , signal proce 1024 point
p plot for a s
ected by the usly describe
DC02 (a = C02 (a = 0
ve been com wo largest va
n in Table 2 ir limits (IAE
values (1st PC01, and P
2nd IAE C
1400 1700 103 147
can see tha value and co ted by perio
on Source
e collected an ampling peri
hat the data hen, each p pwelch funct essing toolbo t for each va
M. Farenzena, C
Brazilian Jou
sticky valve.
e application ed, consider
= 1, Ti = 6 0.5, Ti = 1
mputed for alues (1st IA 2 for these f
ELIM).
IAEC and
PC02. IAELIM 191 191 20 20
at all values onclude that odical load d
nd all measu od (5 second a was collec
ower spectr tion in Matla ox version 4 ariable and
C. A. Kayser and
urnal of Chemica
n of ring 600 120 the AEC four 2nd M ex-the dis-ured ds). cted rum ab® .2). the wi wo nat con al. 4 s the tha com pu fie can loo Fig eac Ta the F F F F F D D L L L L P P P
J. O. Trierweiler
al Engineering
ndow size i ork was 256. The periodi te have perio nds. Using , 2001), the m shows the in e number of
Based on F at should be mposition, th uted (see Tab
ed as PC01 a nce index. Th op is shown i
gure 4: Spec ch plot show
able 3: Signif e sources of
IC1 FC01 0.07 FC02 0.11 FC03 0.02 FC06 0.08 FC08 0.00 DC01 0.02 DC02 0.01 LC01 0.73 LC02 0.67 LC03 0.02 LC04 0.02
PC01 1.00
PC02 0.95
PC04 0.54
in the Fourie
ical disturban ods equal to the FastICA matrices A an ndependent p
ICs in 5. Figure 4, the
eliminated he significan ble 3). The so and PC02, a he impact of in Figure 5.
ctral indepen ws one indepe
ficance matr f IC1 are hig
1 IC2 7 0.01 1 0.05 2 0.01 8 0.01 0 0.03 2 0.03 1 0.02 3 0.21 7 1.00
2 0.02 2 0.00
0 0.29
5 0.47
4 0.07
er Transform
nces that we o 320 second A algorithm
nd S were ob power spectr
e independen is IC1. Base nce matrix w ources of IC as assessed b f each IC ove
ndent compon endent comp
rix for the I ghlighted.
IC3 I
0.07 1
0.55 0
0.00 0
0.01 0
0.21 0
1.00 0
0.98 0
0.00 0
0.00 0
0.89 0
0.77 0
0.01 0
0.10 0
0.10 0
m used in th
want to elim ds and 427 s
(Hyvärinen btained. Figu ra. We limite
nt disturban ed on ICA d was then com C1 were ident
by the signif er each contr
nent analysis ponent.
CA analysis
IC4 IC5 1.00 0.19
0.05 0.82 0.02 0.01 0.09 0.00 0.32 0.22 0.12 0.51 0.12 0.52 0.57 0.13 0.21 0.04 0.17 0.72 0.19 0.91 0.25 0.18 0.07 0.15
0.01 1.00
F p v st ob D ti b o lo hy su p b th A w th
Figure 5: Hi endent com ariable.
Figure 5 s tricted to PC bserved loop
Determining
Subsequen ion, its cause oth valves su First, the P logy based o oop is non-G
ypothesis of ult and quan lot was draw
Fig
Figure 6 c le stiction. T his case, the
In the cas Again, the lo which indicat he ellipsis int
0.00 0.50 1.00 1.50 2.00 2.50 Plantw Brazilian istogram wit mponent (IC)
shows that th C01 and PC0
ps.
the Oscillat
nt to determi e should be uffer from st PC02 loop wi on higher-ord Gaussian and f a sticky va ntify the stic wn, as shown
gure 6: pv ve
clearly indica The sb is 1.
valve should e of PC01, op is also n tes another st
terpolation f
wide Periodical D
Journal of Chem
th the impac ) (significan
he impact of 2. It spreads
tion Cause
ining the sou diagnosed. W iction. ill be analyze der statistics d nonlinear, alve. To corro ction (sb), th n in Figure 6.
ersus op for P
ates that the 18% of the d be replaced the same an on-Gaussian ticky valve. for the pv ver
Disturbances Isola
mical Engineering
ct of each in nce) over ea
f IC1 is not s its effect in
urce of osci We suspect t
ed. The meth shows that t confirming oborate this he pv versus
PC02.
valve has v valve span. d.
nalysis is ma n and nonline
Figure 7 sho rsus op curve I I I I I
ation and Elimina
g Vol. 32, No. 04,
nde-ach re-n all lla-that hod-this the re-op In ade. ear, ows e. is a Ap wa val thi any the ver Th spe the pla We dis sig op Fig pen abl IC5 IC4 IC3 IC2 IC1 0 0 1 1 2 2
ation in a Petroch
, pp. 919 - 927,
Based on Fi again confirm
Figu
pplying the S
Finally, afte as PC01 and lve stiction, is hypothesis
The mainte y valve, bec em. Howeve rify if the os In the first hen the data ectral indep e histogram ant when PC
The same p e open this sappears. Fig gnificance in
ened.
gure 8: Hist ndent compo le with PC01 .00 .50 .00 .50 .00 .50 hemical Unit
October - Decem
igure 7, the h med. The sb f
ure 7: pv vers
Solution in a
er we diagno PC02, and th
we should s through a te enance group cause there i er, we open cillation disa test (Test1), a was collec endent com with the sig 01 was open procedure wa
loop for a gure 9 show ndices for th
togram with onent (IC) (s 1 open (Test1
mber, 2015
hypothesis of for PC01 is e
sus op for PC
a Real Plant
osed that the he cause in b then corrob est in a real p p does not al is no bypass each loop fo appears.
the loop PC ted and dec mponents. Fig
gnificance in n.
as made for period to v ws the histog he plant wh
h the impact significance)
1).
9 f valve stictio equal to 0.90.
C01.
t
source of IC both cases w borate (or no
plant. llow replacin s available f for a period
C01 was ope composed in gure 8 show ndices for th
PC02 (Test2 verify if dis gram with th hen PC02 w
of each ind ) on each var
926 M. Farenzena, C. A. Kayser and J. O. Trierweiler
Brazilian Journal of Chemical Engineering
Figure 9: Histogram with the impact of each inde-pendent component (IC) (significance) on each vari-able with PC02 open (Test2).
Based on Figure 8 and Figure 9, we can verify that, in each case, the disturbance vanished. The positive impact on loop variability is corroborated by the comparison between the ratio of the original (when all loops were closed) and the new variance when PC01 and PC02 were opened, as shown in Table 4.
Table 4 summarizes the strong impact caused by IC1. Its elimination can reduce the product and the whole plant variability, achieving a more profitable operating point.
Table 4: Ratio between the original and the new variance, when PC01 and PC02 were opened (see that the variance of same loop was normalized using the same parameter, for the three cases).
Loop Normalized original variance
Relative variance - PC01 open
Relative variance - PC02 open
DC01 1 -78% -48%
DC02 1 -88% -93%
FC01 1 -98% -94%
FC02 1 -83% -84%
FC03 1 -30% -91%
FC06 1 1724% 2%
FC08 1 86% -78%
LC01 1 -26% -15%
LC02 1 -1% 3%
LC03 1 1400% 2035%
PC01 1 171% -67%
PC02 1 -51% 6867%
PC04 1 98% -29%
CONCLUSIONS
In plants with a high number of recycles, periodi-cal disturbances can strongly affect loop variability, because one oscillating loop can have a widespread effect in the whole plant. Thus, isolating its source and diagnosing its cause are essential to ensure a
highly efficient operation. In this work, we illustrate this scenario in a petrochemical plant located in Southern Brazil, where periodical disturbances have a strong influence on product variability.
The procedure followed three steps. Initially, we detected loop oscillation using the methodology based on IAE (Hägglund, 1995). Then, each disturb-ance was isolated using Spectral Independent Com-ponent Analysis (Xia et al., 2005). We identified the disturbance whose period we want to isolate and its sources, which were PC01 and PC02. Finally, we diagnosed the cause of the fault in both loops as valve stiction using the methodology based on higher-order statistics (Choudhury et al., 2004).
To verify the theoretical predictions in the indus-trial plant, each loop was opened for a period, be-cause the valves could not be replaced. The impact was visible: the reduction in the variability of almost all loops was verified and the reduction in product variability was up to 93%.
ACKNOWLEDGMENTS
The authors would like to thank CAPES, PETRO-BRAS and FINEP for supporting this work and the petrochemical company that kindly provided the data.
REFERENCES
Brásio, A. S. R., Romanenko, A., Fernandes, N. C. P., Modeling, detection and quantification, and compensation of stiction in control loops: The state of the art. Industrial and Engineering Chem-istry Research, 53(39), 15020-15040 (2014). Choudhury, M. A. A. S., Shah, S. L. and Thornhill, N.
F., Diagnosis of poor control loop performance using higher-order statistics. Automatica, 40(10), 1719-1728 (2004).
Choudhury, M. A. A. S., Shah, S. L. and Thornhill, N. F., Diagnosis of Process Nonlinearities and Valve Stiction: Data Driven Approaches. Berlin, Springer (2008).
Choudhury, M. A. A. S., Shah, S. L., Thornhill, N. F. and Shook, D. S., Automatic detection and quan-tification of stiction in control valves. Control En-gineering Practice, 14(12), 1395-1412 (2006). Duan, P., Chen, T., Shah, S. L., Yang, F., Methods for
root cause diagnosis of plant-wide oscillations. AIChE Journal, 60(6), 2019-2034 (2014).
Hägglund, T., A control-loop performance monitor. Control Engineering Practice, 3(11), 1543-1551 (1995).
0.00 0.50 1.00 1.50 2.00 2.50 3.00
Plantwide Periodical Disturbances Isolation and Elimination in a Petrochemical Unit 927
Brazilian Journal of Chemical Engineering Vol. 32, No. 04, pp. 919 - 927, October - December, 2015
Horch, A., A simple method for detection of stiction in control valves. Control Engineering Practice, 7(10), 1221-1231 (1999).
Hyvärinen, A., Karhunen, J. and Oja, E., Independent Component Analysis. New York, J. Wiley (2001). Jiang, H., Shoukat Choudhury, M. A. A. and Shah, S.
L., Detection and diagnosis of plant-wide oscilla-tions from industrial data using the spectral enve-lope method. Journal of Process Control, 17(2), 143-155 (2007).
Oppenheim, A. V., Schafer, R. W. and Buck, J. R., Discrete-Time Signal Processing, Upper Saddle River. N. J., Prentice Hall (1999).
Rossi, M. and Scali, C., A comparison of techniques for automatic detection of stiction: Simulation and application to industrial data. Journal of Pro-cess Control, 15(5), 505-514 (2005).
Singhal, A. and Salsbury, T. I., A simple method for detecting valve stiction in oscillating control loops. Journal of Process Control, 15(4), 371-382 (2005).
Thornhill, N. F., Finding the source of nonlinearity in a process with plant-wide oscillation. IEEE Transactions on Control Systems Technology,
13(3), 434-443 (2005).
Thornhill, N. F., Cox, J. W. and Paulonis, M. A., Diagnosis of plant-wide oscillation through data-driven analysis and process understanding. Control Engineering Practice, 11(12), 1481-1490 (2003).
Thornhill, N. F. and Hägglund, T., Detection and di-agnosis of oscillation in control loops. Control Engineering Practice, 5(10), 1343-1354 (1997). Thornhill, N. F. and Horch, A., Advances and new
directions in plant-wide disturbance detection and diagnosis. Control Engineering Practice, 15(10), 1196-1206 (2007).
Xia, C. and Howell, J., Isolating multiple sources of plant-wide oscillations via independent component analysis. Control Engineering Practice, 13(8), 1027-1035 (2005).
Xia, C., Howell, J. and Thornhill, N. F., Detecting and isolating multiple plant-wide oscillations via spectral independent component analysis. Auto-matica, 41(12), 2067-2075 (2005).
