industrial applications of predictive functional control

End of the Sixties:
- ORIGIN OF MODEL BASED PREDICTIVE CONTROL
INDUSTRIAL APPLICATIONS
OF
PREDICTIVE FUNCTIONAL CONTROL
- PETROLEUM INDUSTRY / DEFENCE INDUSTRY
.Refineries :
How to handle constraints on MV’s and CV’s ?
How to control a multivariable process ?
to our dear friend :
David CLARKE !
.Defence :
How to control a process with a non stationary set point
with no lag error (follow-up servo) ?
Oxford 09/01/2009
Jacques Richalet
jacques. richalet @ wanadoo.fr
WHY
WHYMPC
MPC ??
PROFIT
OPTIMIZATION
CONSTRAINTS*
Gasoil
Crude oil
PREDICTION
MODEL
Residue
M.B.P.C / PREDICTIVE CONTROL
Tracking
Tracking servo
servo system
system
CONTROL
CONTROL ON
ON CONSTRAINT
CONSTRAINT
Ko
Viscosity
of residue
LIMIT !
Ok
H
O
t
1975:Total Normandy Refinery. D11column 30.000
T/day!
D
….the past…
NO
NO LAG
LAGERROR
ERROR
1968 : Basic principles of Predictive Control
1973 : 1st version of PFC/ IDCOM
1974 : 1st industrial applications of PFC/IDCOM
P
X
Steam Generator/ Reactor/ Distillation
εΤ = 0 !
1978 : 1st contribution MPC “Automatica “ ( problems..!)
1980-85 : HIECON with Set Point ( Houston)
Target
trajectory
1989 : Extension from PFC to PPC:
Heading
t
Parametric Predictive Control (control by reactor flow..)
1998 :1st implementation of an MPC in the native library of a
PLC : Modicon. Schneider-Electric.Momentum.Concept
FIRST
FIRST INDUSTRIAL
INDUSTRIAL APPLICATIONS
APPLICATIONS OF
OF P.F.C.
P.F.C.
DEFENCE
Mine hunter
Ariane 5 attitude control
Missile autopilot
Laser guided missile (t=62 microsecond…)
Gun turrets
Radar antennas
Infra Red camera
High speed Infra Red
Missile launch
Camera mount
Radar antennas
Laser mirror
Tank turret (T. 55)
Mine sweeper auto-pilot
Aircraft carrier auto-pilot
AUTOMOTIVE
Gear box test bench
Dynamic test bench engine
Fuel injection
Idle fuel injection
Clutch antistroke
Gear box (tank)
Hybrid car (electric-fuel)
Air conditioning
METALLURGICAL INDUSTRIES
Coating lines aluminium
Thickness control - Roll eccentricity
Mono-multi-stands Rolling mills
Continuous casting (slab)
Push-ovens
Coke furnace
Hot/cold/Thin rolling mills
Steam generators
Level, temperature, pressure……;
MISCELLANEOUS (Chem reactors and
distillation excluded)
Bioreactor Melissa ESA
Plastic extrusion robot (high speed)
Temperature control of gas furnace
Temperature control of TGV train carriage
River dam level control (T=1 hour )
Powder milk dryer
Electric furnace brazing etc….
FEATURES
EXOTIC APPLICATIONS…
North : Molde (Norway)
South : Plaza Huincul ( Patagonia-Argentina)
East : Hokaido ( Japan)
West : Mobile
( USA.DEGUSSA)
High : Eiffel Tower elevator !
Space : European Space Agency : Mars project bioreactors ( 2029 !)
Depth : Mine hunter boat
Speed : Temperature cabin ( TGV : 574.8 km/h W.record )
Fast : Laser guided bomb ( Tsampling : 65 s )
Slow : River dam level
( Tsampling : 1 hour)
Ecology: Diester from colza
Animal : Dog food pellet dryer
Plants : Greenhouses
etc…
…the oldest MBPC… 1968
A “few” thousand applications in many fields
AT WHAT LEVEL DO WE OPERATE ?
N Histogram
2
Back to Basics:
• Level 0 : Ancillary processes e.g: FRC/ Pid
(the valve is the nightmare of control !)
• Level 1 : Dynamic control with constraints
• Level 2 : Optimization of working conditions
• Level 3 : Production planning
OFF
SPEC
1.5
1
Sigma 1
0.5
.Level N is operative if level N-1 works well…!
Q1
. »There are no technical problems, but only technical
aspects of economic problems … »
0
0
50
100
150
200
QUALITY
250
300
350
400
450
N Histogram
N Histogram
LEVEL 0 / LEVEL 1
2
LEVEL 0 / LEVEL 1
2
Sigma 1 -- Sigma 2
Sigma 1 -- Sigma 2
Sigma 2
Sigma 2
OFF
SPEC
1.5
1.5
1
1
Sigma 1
Sigma 1
0.5
0.5
Q1
0
LEVEL 2
Q1 -- Q2
OFF
SPEC
0
50
100
150
QUALITY
200
« SQUEEZE
250
AND
300
350
400
Q2
0
450
0
50
100
150
Q1
200
QUALITY
250
300
350
400
450
SHIFT »
44 BASIC
BASICPRINCIPLES
PRINCIPLESOF
OFPFC
PFC ::J.
J.PIAGET
PIAGET
C OS T FUNC TION
N Histogram
LE V E L 0 / LE V E L 1
2
S igma 1 -- S igma 2
S igma 2
OFF
SPEC
– Operating Image
- Internal independant Model
– Target – Sub Target
– Reference Trajectory
– Action
– Solver. Functional basis
Structured future MV
LE V E L 2
Q1 -- Q2
W2
1.5
D E LTA C OS T (W2-W1)
W1
1
S igm a 1
0.5
– Comparison :
– Predicted / Actual
Q2
0
0
50
100
150
Q1
200
– Error compensator
QUA LITY
250
300
350
400
450
STRATEGY…
. Control of industrial processes : PI(D) ?
but PID cannot solve all control problems,
but any candidate controller should « look like a PId »:
EASY TO UNDERSTAND, TO IMPLEMENT, TO TUNE
.. consistent with floor instrumentists’ habits
so:
. Tuning parameters should have a clear, physical interpretation.
. Elementary mathematics.
. No explicit integrator in the loop.
. No matrix calculation.
. No quadratic minimization on-line.
. No iterative computation on-line.
. Open technology…
Natural Control : “You would not drive your car using a PID scheme”
•
TARGETED PERFORMANCES
. Control « all » types of processes: time delay, unstable,
non-minimum phase, some non-linear
. Handles all constraints on the MV and on internal
variables of the process CV.
. No lag error on dynamic set points with no integrator
Industrial settings:
. Transparent and Cascade Control with transfer of
constraints from inner loop to outer controller (« Back
Calculation »)
. True feed-forward
. Split-range control with different dynamics (reactors!)
. Control with 2 cooperative MV’s, e.g., big valve / small
valve
. Extendable to 2 MV/ 2 CV…..
. Dual control: total elimination of harmonic disturbances
. Full Robustness analysis: easy trade-off ,
.
Immediate tuning
/
Open technology
2
n
Past
REFERENCE
REFERENCE
TRAJECTORY
TRAJECTORY
Future
ε (n+H)
Set point
ε (n)
Y
P
E
S
O
F
M
O
D
E
L
S
C(n+H)
perturbation
Reference
trajectory
Predicted output
y
T
TRBF
perturbation
y ref
∆
P
e
+
+
Process
sp
Model
sm
e
+
+
Process
sp
Model
sm
Process output
yP
∆ M
Re-aligned Model
Model output
Independent Model
H
H 1
yM
H 2
Independent models are selected:
Coïncidence
horizon
. Input /Output models valid for all math. structures
MV
Manipulated variable
MV i
Prediction
horizon
Solver
. No permanent errors in steady-state modes!
( never mix process variables with model parameters..)
Why structuring the future MV?
• The future MV is projected on a Functional Basis:
• Determing all future MVs is of little benefit:
– many projects lead to the same future behaviour
– e.g. Taylor expansion / polynomials
• (low-pass process)
• Thus, the MV is restricted a priori and not damped
afterwards….
j. Uj(i)
• MV(n+i) =
• U0(i) = 1, U1(i) = i U2(i)= i2 etc …
• Find the j such that:
Σ
– At a finite number of points, the predicted model output
coincides with the desired reference trajectory
• Computation time increases ( PLC !)
• Introduction of a damping term on the speed
of the MV a priori is difficult to tune !
• Projection on « Eigen » functions of linear
dynamic processes insures no lag error on
dynamic set points
• (i.e., coincidence points)
ClOSED LOOP WITH CONSTRAINT
OPEN LOOP
140
140
120
120
MV CONSTRAINT
SETPOINT
MV
100
100
80
80
CV
h=125
60
exp(−20s).(1 − 15s)
H (s ) =
(1 + 25s)(1 + 55s)(1 + 80s)
CLTR DESIRED =410
CV
60
CLTR ACTUAL =424
40
40
dCV
20
20
0
0
-20
0
100
200
300
400
500
600
700
800
900
0
100
200
300
400
500
time
600
700
800
900
REFERENCE TRAJECTORY at ti=20
REFERENCE TRAJECTORY at ti=60
140
140
MV
MV
120
120
Setpoint=100
Setpoint=100
100
100
80
80
60
60
REF TRAJ.(green)
REF TRAJ.(green)
CV PREDICTED(black)
40
CV PREDICTED(black)
40
CV PASSED(red)
20
CV PASSED(red)
20
tinit=20
tinit=60
0
0
-20
100
200
300
400
500
600
700
800
900
-20
100
200
REFERENCE TRAJECTORY at ti=100
300
400
500
600
700
800
900
800
900
800
900
REFERENCE TRAJECTORY at ti=150
140
140
MV
MV
120
120
Setpoint=100
Setpoint=100
100
100
80
80
60
60
REF TRAJ.(green)
REF TRAJ.(green)
CV PREDICTED(black)
40
CV PREDICTED(black)
tinit=150
40
CV PASSED(red)
20
CV PASSED(red)
20
tinit=100
0
0
-20
100
200
300
400
500
600
700
800
900
-20
100
200
REFERENCE TRAJECTORY at ti=200
300
400
500
600
700
REFERENCE TRAJECTORY at ti=300
140
140
MV
MV
120
120
Setpoint=100
Setpoint=100
100
100
80
80
tinit=300
tinit=200
60
60
REF TRAJ.(green)
REF TRAJ.(green)
CV PREDICTED(black)
40
CV PREDICTED(black)
40
CV PASSED(red)
20
20
0
0
-20
100
200
300
400
500
600
700
800
900
-20
CV PASSED(red)
100
200
300
400
500
600
700
ACCURACY
ELEMENTARY
ELEMENTARYTUTORIAL
TUTORIALEXAMPLE
EXAMPLE
• Basic Eigen Function property:
• If the set point can be projected on a polynomial basis
of order N, and if the non integrative model is of order
N, then there is no lag error
• Accuracy depends on the choice of the functional basis
and not on the gain of the controller….
P=
y
G e-θ s
=M=
1 + Ts
u
θ = Tsamp . r
Trajectory expo.
y(n) = α y(n+1) + (1- α) . u(n-1-r) . G α = e
-Tsamp
T
Processus with delay yPr (n) : = yP (n) + yM (n) - yM(n-r)
Target
:λ
1 coincidence point: H
1 Base function: step
∆P(n+H) = (C - yPr ) (1 - λH)
:
Model
:
Model increment :
y M (n) α H
Free mode
Increment = Free(n+h) + Forced(n+h)- ymodel(n)
Forced mode u(n) . GM  1 - α H 
The only mathematical problem for trainees …!


:
Control equation :
∆ P (n + H) = ∆ M (n + H)
(C - y
(n )) (1 - λ H ) = y M (n ) α H + u (n ) G M (1 - α H ) - y M (n )
Pr
u(n) =
(C - y (n)) (1 - λ H )
y (n)
Pr
+ M
GM
G M (1 - α H )
SENSITIVITY / CLTR
0.4
DUAL CONTROL
LOG(CV/DV)
0.2
0
• Problem:
-0.2
– Classical control:
• Constant set-point + harmonic disturbance rejection
. Complex algebra controller : much simpler !
• Examples:
-0.4
CLTR=300
-0.6
-0.8
– Helicopter : 20mm artillery / vibration damping
– Continuous casting steel industry :
oscillation of the mould and roll swell effect
– Wave Energy Converter etc..
CLTR=30
-1
-1.2
-1.4
LOG(omega)
-1.6
-6.5
-6
-5.5
-5
-4.5
-4
-3.5
-3
-2.5
-2
Mod|CV/DV| b withoutDual / r with Dualcontrol
Temperature control with internal variable constraints
0.25
tauprocess=50
damping
0.2
P
e
l
e
c
omeg=1 // tau =150)
T
0.15
a
i
r
CLTR=200
without dual
with dual
T
0.1
Tsurf
0.05
wave pulsation 1=2pi/150
0
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6
1.8
i
n
t
SETPOINT CV1=100 / CONSTRAINT CV2 = 129
Constraints on internal variables:
the multiple controller approach !
SM2
160
MV1
140
M2
CONSTRAINT CV2
120
CV2
CV2
Cons2=CV2max
R2
contrainte
constraint
P2
CV2
(b)
100
80
(a)
t
MV2
Superviseur
Supervisor
R1
MV1
Cons1
Cons1
20
MV2 active
MV1 active
CV1
SM1
CV1
40
CV1
P1
MV2
60
0
PERT
t
M1
-20
Extension to different dynamic constraints …
100
200
300
400
500
600
700
800
TUNING
GAIN MARGIN / CLTR
7
Base
Function
GAIN MARGIN
6.5
6
5.5
CLTR Horizon
Accuracy
2
0
0
Dynamics
0
2
1-
Robustness
0
1+
2
5
4.5
OLTR
4
20
40
60
80
100
CLTR
120
140
160
Goodwin ?
Types of model:
1)
Black Box
• Physical analysis . Choice of model ?
• Application of a Test signal protocol ?.
• Optimisation of the protocol?
• TOUGH CONSTRAINTS from THE PRODUCER.!
• Pseudo-random binary noise to be avoided in practice …
• Action on-site and off-site : Cost / Man Hours
• No trade ! : to be repeated on all processes
• Weak capitalisation.
• BUT: no investment, easy to perform….
Identification of Industrial Processes
The target is not to minimize the quadratic error
between the process output yp(n) and the model
output ym(n):
Cstate = ∑ ( yp(n)-ym(n) )2
but to minimize the Structural Distance between the
process parameters Pp(j) and the model parameters
Pm(j) ( Lyapunof function !..)
Cstructure= ∑ ( Pp(j)-Pm(j) )2
because Cstate depends on the input of the process
and can be small with wrong parameters so:
• Thus the problem is :
• How to define the most pertinent input signal
leading to the minimum uncertainty of the
parameters taking into account the industrial
constraints ? :
ISO-DISTANCE GAIN / TAU
1.6
1.4
1.2
GAIN
• Limited duration of test protocole: Hmax
H=2000
1
H=400
• Limited amplitude of test signals : Amax
H=150
0.8
• Limited frequency spectrum
:
ω
< max
ω
0.6
• The design of the « best protocol» becomes, mainly,
deterministic : iterative learning procedure !
H=50
0.4
0
10
20
2) First principles models
: knowledge based model
• Physical analysis of the process:
• « Brain juice »:
– Tough - expensive – risky ...
– Expertise needed in Physics, Chemical engineering
,Mechanics, etc.
• but ….
– Generic. Fewer calendar days and man-hours
– PHYSICAL ADAPTATION OF THE CONTROLLER
– Strong capitalisation
the future………
30
40
50
Taum
70
80
90
Transfer / Training
• Vendors :
–
–
–
–
–
–
–
–
–
–
SET POINT (USA Houston)
P.Latour / P.Grosdidier / Tom Badgwell / Greg Martin
. YOKOGAWA (Japon. Mitaka)
. SOTEICA (Argentina / USA)
. SAMSUNG (Korea)
. REPSOL (Spain)
. JGC (Japon / Yokohama)
. SCHNEIDER ELECTRIC (France )
. INSTITUT TECH NAGOYA (Japan)
. UNIV . of Hanghzou (China)
Users….
•
•
•
•
•
•
•
•
•
•
•
•
•
•
60
AMOCO ( CAN)
MOBIL OIL ( USA / F)
TOTAL ( F)
ATO CHEM ( F)
LUBRIZOL ( USA. F)
SOFIPROTEOL( B. F)
BASF ( B. D)
DEGUSSA. EVONIK.( D )
PECHINEY/ALCAN (F )
ARCELOR-MITTAL (L )
SANOFI- AVENTIS ( F.D )
IRA (F) etc….
VEOLIA (F)
..and others..
PID + LOGIC
100
Péchiney Aluminium 1988
Cold Rolling Mills : 16 / 18
Thickness control Ts =5ms
Flatness control
Ts=20ms
F.P. models : Self.tuning = speed/ thickness/ metal
+ Low level on line identification of model (lubrification !)
Speed
PFC without Logic
Stdev
PID
250 m/mn
4.5 µm
PFC
1450 m/mn
0.7 µm
Pay Out Time ?
Fichier nivocc1.mat
nivocc1 MV(k) CV(r) Cons(b)
85
80
ident. 2
75
70
65
60
ident. 1
7000
ISO-DISTANCES IN THE PARAMETRIC SPACE
8000
9000
10000
sec
11000
12000
13000
Comparaison PID / PFC
nivocc1 ISO-D
5.5
5.4
Ident. 2
Mesure du niveau d’acier (70%)
5.3
5.2
K
5.1
R=11
5
Mesure position du vérin ~88 mm
Ident. 1
4.9
Mesure de la vitesse ~1.3 m/mn
4.8
4.7
R=11
4.6
4.5
0
0.05
0.1
0.15
0.2
tau
0.25
0.3
0.35
4 Steam Generators
La centrale : une chaudière
Gaz de haut fourneau enrichi
Gaz de four à coke
Fioul lourd
Goudron
1
2
3
4
4 x 215 t/h
80 bars et 500 °C
TA 1
52 MW
TS
70 t/h
prélèvement
Détente 2
Vapeur moyenne pression
13 bars 250 °C
TS
70 t/h
VENT
270 000 M3/H
Utilités
usine
HF
40
Superheater Temperature - Cascade Control
Qdes1+360(b)
378
Qdes t/h
376
Détente 1
t30095a-Constdes1(k) Tdes1(r) tsur1(m)
TA 2
52 MW
TA 4
32 MW
HF
39
Set Point Protocol for Tdes (1GV)
prélèvement
VENT
270 000 M3/H
TA 3
32 MW
PFC2
PFC1
h =(Tm1+Tm2+Tm3)*0.3 (=30.3 s)
Tech=2s
TRBF1=150s
h=2 s
Tech=2s
Le FRC est supposé parfait :
Kp=1.35
Ti=100s.
Zone : largeur = ±5 °C
TRBF2 L = 35 s
TRBF2 H =(Tm1+Tm2+Tm3)*3+retard (=350 s)
MI du 3 ème ordre
nov 05
Charge calorifique
kd=a-b*qvap+c*qvap²
Charg (%)
374
Qvap (t/h)
Tsur (°C)
372
MI :K=1 T=5s
Tdes °C
370
d/dt=50/2 °C/s
d/dt=20 t/h/s
Constvap
500 °C
368
PFC2
366
kd
gdes gdes
~1
0-40 t/h
Ctdes 410°C
PFC1
x
FRC
rcpy
300°C
rcpy
Cqdes
Contraintes
1
1+ Tp
-
tvapint
1
1 + T1 .p
Gch.e − Rch.p
+
G.e-R3.p
1 + T3.p
+
T1=21.5s
Transf. de
contraintes
364
tdes
+
Cqdes*
Procédé Tdes
Procédé Tvap
Régulateurs
362
Tdes en boucle fermée : cte de temps = T2
1000 1500 2000 2500 3000 3500 4000 4500 5000
sec
PFC Histogram
PID Histogram
tvap (°C) et Qvap (t/h)
GV-TEMPERATURE VAPEUR en PFC- Consigne:b - Tvap:r - Qvap*2:k t22043a.mat
600
500
400
300
200
sec
GV-TEMPERATURE VAPEUR en PID- Consigne:b - Tvap:r - Qvap*2:k t22043a.mat
550
tvap (°C) et Qvap (t/h)
500
500
450
400
350
300
0
1000
2000
3000
4000
5000
6000
7000
8000
0
1000
8
Frequence en %
Frequence en %
3000
4000
5000
6000
sec
7000
8000
10
10
Tvapmoyen=499.7504
ecart-type=2.352
Min/maxTvap=492.328 507.983
tmin/tmax (s)=3000 7000
6
4
2
0
-30
2000
Histogramme de l'écart TVAP - t22043a
Histogramme de l'écart TVAP - t22043a
-20
-10
0
10
ECART TVAP en °C
20
30
8
Tvapmoyen=488.2418
ecart-type=3.7752
Min/maxTvap=478.784 500.757
tmin/tmax (s)=3000 7000
6
4
2
0
-30
-20
-10
0
10
ECART TVAP en °C
20
30
ECONOMIC EVALUATION
• Heating / Ventilation / Air-conditioning
Efficiency of turbines increases with steam temperature
• Europe : 10% increase 2006 - 2007
Above 519 °C the blades of the front compressor starts
Creeping ……. and hit the case
• Market : 14 MM€
Classical problem : « 2 S’s »
« Squeeze the variance and Shift the target »
3°C
--------------
• Energy saving / Global warming
1M€
• Ease of implementation
Original set point : 493°C -- Today: 506°C
Pay Out Time : meaningless…!
CLASSICAL COOLING PROCESS: Mollier diagram
4.1 Moyens mis en oeuvre
LOAD CHANGE
PID
140
HP
15
120
100
T°C
80
5
!
Danger !!!!
OVERHEATING
60
0
40
-5
20
-10
9:36
0
10:04
10:33
Surchauffe
11:02
Consigne
11:31
Temps
Puissance Electrique
12:00
P
12:28
12:57
Platelec
LOAD CHANGE
PFC
140,00
15,00
120,00
T°C
80,00
5,00
60,00
Danger!
0,00
40,00
-5,00
20,00
-10,00
10:48
0,00
11:16
11:45
12:14
12:43
13:12
Temps
Surchauffe
Consigne
Puissance Electrique
P
Platelec
13:40
14:09
kWe Platelec
100,00
10,00
Model Predictive Control and Realtime Optimization Implementation
Policy and Best Practice
( from ARC )
kWe Platelec
10
Do you see your use of MPC accelerating, staying constant, or declining?
Is it your company’s objective to implement and support your MPC yourself or outsource?
97 Increasing
40 Staying Constant
2 Decreasing
1.4 %
Decreasing
112 In-house Implementation and Support
28 Third Party Please specify
80.0 %
120
110
100
90
80
70
28.8 %
Staying Constant
60
50
20.0 %
40
30
69.8 %
Increasing
20
10
0
20
40
60
80
0
In-house Implementation and Support
100
Third Party Please specify
Vertical
33 Refining
28 Chemicals
21 Oil & Gas
15 Other
7 Pulp & Paper
7 Power
7 Electronics
4 Plastics & Rubber
3 Metals
3 Cement & Glass
2 Mining
2 Automotive
1 Aerospace
1 Food & Beverage
1 Machinery
1 Pharmaceuticals
0.7 %
0.7 %
0.7 %
0.7 %
1.5 %
1.5 %
2.2 %
2.2 %
2.9 %
Pharmaceuticals
Machinery
Food & Beverage
Aerospace
Automotive
Mining
Cement & Glass
Metals
Plastics & Rubber
Electronics
Power
Pulp & Paper
Other
Oil & Gas
Chemicals
Refining
5.1 %
5.1 %
5.1 %
11.0 %
15.4 %
20.6 %
24.3 %
0
5
10
15
20
25
30
35
Why Chemistry lags behind Petroleum ?
a) The higher the added value: the lower the interest in
production improvement.!
b) Batch reactor specific problem:
Cooling fluid unstable but Mass Temp. stable !
c)
PFC/PPC
in the
Chemical Industry
Gap between :
Chemical Engineering / Automatic Control
« School reproduces Industry / Industry reproduces School »
or
: « heat exchange based industry »
.. the trend…!
Constraints :
– Oil barrel price is uncertain.
– Lethal Asian competition.
– Quality norms (FDA)
– Ecology
But fortunately :
– Controlers and software are more accessible
– Incoming of a new generation of personnel
– The « crisis » will force Industry to improve!
Convexity Therorem
4 CONTROL STRATEGIES OF BATCH REACTORS
F1, T1
R Reagent
R
M
TM
F i , Ti
∞
Te
dT
M
= UA T - T + ∆ Hx
e
M
dt
ρ e CeP Ve dTe = ρ e Fi T i - T e + UA T M - Te
dt
.
θ(Fi) TM+ TM= Ti
ρ CM V
P
M
1) Fi =ct /MV= Ti CV=TM / level 0 =Ti ?
:PFC
2) Ti =ct (!) MV=Fi Parametric Control non linear :PPC
3) MV : Ti and Fi : Enthaplic control (power)
:PPC+
4) MV : Pressure of reactor
:PFC
75
F, T
F2, T2
F.T =F1.T1+F2.T2
T=λ
λ.T1 +(1- λ).T2
λ = F1/(F1+F2)
0≤λ≤1
tsf
Equivalent Temperature Procedure
tsp
• . dT/dt +T = T1+( 1 - ) T2
• . dT/dt +T = TEQ
• PFC computes TEQsol ( sol)
• sol is connected to the physical MV by
a non-linear relation
• Non-linear solver outside the dynamics!
qp, tep
qf, tef
1
1
− )]
Fp Ff
Γ(Qf ) =
Fp
1
1
1−
exp[−U.A( − )]
Ff
Fp Ff
1 − exp[−U.A(
Fp, Ff : Thermal flows
Fp = (ρ
ρ.Cp)p.Qp
Ff = (ρ
ρ.Cp)f.Qf.
Physical Model: Heat Exchanger/ Reactor
• Mass balance/ Enthalpy balance
• Parameters : Geometry / Fluid
Characteristics / Flows and Temp
• Heat exchange coeff. U : Wm2/K as a
function of Reynolds/ Prandtl/ Nusselt
• Function of temperature and flows
NATURAL PHYSICAL SELF ADAPTATION OF
THE INTERNAL MODELS OF PFC
60
55
TEP d°C
50
45
QP L/H/5
40
35
TSP d°C
30
25
20
QL L/H/5
15
10
0
100
200
300
400
500
600
700
TEMPERATURE DE MASSE
120
Tmax=108.4°
d°
100
92°
80
60
Tinit=42°
40
20
0
min
0
20
40
60
Figure 4
80
100
120
EXOTHERMICITY ESTIMATOR
by INVERSE PFC
B
Pert
Cons1
+
+
MV1
R1
P
+
MV2
SP
+
SP*
M≡
≡P
R2
Cons2=SP
STRUCTURE and STATE DISTURBANCES
Pert*
+
• On line estimator: an example!
DV (n) = actual disturbance
DV*(n) = estimated disturbance
Gp = 1/ gain of process
Gm = 1/ gain of model (same dynamics…)
DV* (n)= DV(n) +Setpoint.( Gm-Gp )
State and structure mismatches in the same
equation !
E S T IM A T IO N D E L ’E X O T H E R M IC IT E
120
100
T2
T m asse/ Tm asse
e s t im é e
Rea cta nt
Flow
80
Steam
T3
C ooler
60
Heat
Exchanger
T4
E X O T H E R M IC
IT E
40
TS
20
T1 Reactor
TSTE
0
m in
0
10
20
30
40
50
60
70
80
BASF
Figure 3
BASF
31
BASF
BASF
DEGUSSA. EVONIK
World n°1 Specialty Chemistry
n° 3 German Chemical industry
49.000 employees
PFC / PPC at DEGUSSA
• Batch reactors:Heat exchangers
• Esterification : Batch time reduction : approx. 10%
• Polymerization: Temperature control error / 10
• Continuous reactors
> 100 production sites WW
Polymerization . Improved reproductibility. Quality
Technical Center Hanau ( Hesse D.)
Other Units:
Adopts Predictive Control PFC in 2003
Increase of production . Less energy consumption
-----------------------------------------------------------------------
Training of Control Staff
Standardization of software and procedures
Control Systems
INVENSYS PLS 80E (ex Eckhart)
FOXBORO IAS ( HLBL..!)
SIEMENS PLCS7
ABB Freelance 2000
1/01/2009 : > 15 sites WW / 22 workshops / >180 PFC/PPC
8 engineers : work load..?!
ECONOMIC IMPACT…
- Increase of production
- Safety . Repeatability
- No more off-specification products
- Economy : Energy / Solvant
YOKOGAWA DCS 3000
- Maintenance cost lower
EMERSON DeltaV
- Operators : happy : less work load !
In all DCS PFC/PPC is implemented with a generic procedure
: « Structured Text » but mainly by « Block programming »
(IEC1131.3) compatible with Simulink !
Pay out time : « very short »
-Conclusion : large dissemination…
-------------------------But where are the problems ?.......;
Anwendungsbeispiel 1:
Exotherme Batchreaktion
MPC-Regler
für Temperatur
Neue Reglerstruktur:
TC
Standard-Regler
für Kühlwassermenge
FC
TI
TI
04. April 2006
Dr.-Ing. Wolfgang Deis (S-TE-VTA-I)
97
Anwendungsbeispiel 1:
Exotherme Batchreaktion
Control structure
τ=
Entwurf, Simulation und Vorparametrierung in Matlab
(Auszug):
a + b ⋅Qf
Qf
Teq*
TM,Stp PFC
Teq
Qf,Stp
τ
V
TM
PPC Qf,Stp = f(τ
PID
plant
f(τ)
Qf
τ* = f-1(Qf)
trbfPFC = trbfPPC / 3 : Back calculation of constraint
04. April 2006
Dr.-Ing. Wolfgang Deis (S-TE-VTA-I)
99
Anwendungsbeispiel 1:
Exotherme Batchreaktion
85
Anwendungsbeispiel 1:
Exotherme Batchreaktion
85
alte Fahrweise (PI-Regler)
neue Fahrweise (MPC-Regler)
80
+/- 0,2 °C
Temperatur [°C]
Temperatur [°C]
80
75
+/- 2 °C
75
70
70
hier liegen 35 Batches übereinander !
65
0
04. April 2006
20
40
60
Zeit [min]
Fortschritt [%]
Dr.-Ing. Wolfgang
Deis (S-TE-VTA-I)
80
100
120
101
65
0
20
40
60
Zeit [min][%]
Fortschritt
80
100
120
PID
Anwendungsbeispiel 2:
Endotherme Batchreaktion
T REACTOR SETPO INT
T REACTOR [%]
89
[%]
120
100
88
80
87
60
40
0
50
100
150
200
250
300
350
0
50
100
150
200
250
300
350
0
50
100
150
200
250
300
400
120
20
100
0
0.2
0.4
0. 6
0.8
1
1.2
[%]
P ID: Reaktortemperatur [% des Sollwertes]
M PC: Reaktor temperatur [ % des Sollwertes ]
0
80
1.4
MV [%]
60
120
400
82
80
80
78
[%]
100
60
76
74
40
72
LOAD [%]
20
350
400
time [%]
P ID: Dam pfventil [%]
M PC : Dampf ventil [% ]
0
0
0.2
0.4
0. 6
0.8
1
1.2
1.4
• Identification
PFC
89
81
[%]
mass [%]
82
80
88
T REACTO R SETPO INT
T REACTO R [%]
LOAD
0
50
100
150
200
250
300
87
145
350
100
155
160
165
170
175
180
185
150
155
160
165
170
175
180
185
150
155
160
165
170
175
180
190
120
T INPUT REACTO R [%]
MV [%]
80
100
80
60
MV [%]
0
50
100
150
200
250
300
60
145
350
90
89
88
80
78
76
74
87
190
82
T REACTOR [%]
T MODEL [%]
[%]
T [%]
150
[%]
T [°C] PUMPSIGNAL [%]
79
0
50
100
150
200
250
300
350
72
145
Zeit [min.]
LOAD [%]
185
time [%]
Enthalpic Control
Reaching the temp. plateau
SANOFI-AVENTIS
•
•
•
•
•
•
•
Pharmaceutical industry
World n°3 / Continental Europe n°1
World n°1 :Vaccines
Personnel > 140.000
>100 production sites
Turnover : 29MM euros
Drugs, vaccines…
190
SANOFI AVENTIS
PREDICTIVE CONTROL
Identification Mass Temperature
P RE M IE RE ID E NTIF C A TIO N TM A S S E
1 00
Target:
– Improvement of quality of product and repeatability
of recipes. Energy consumption.
– Start : November 2005 .Training of personnel:PFC
- Flexible reactor of the IRA institute in ARLES
•
90
80
Te m p entré e c alo ° /b
70
Tem p .so rtie ca lo e stim ée ! ?/g
60
Controllers:
- Premium Unity Schneider-Electric
50
Tm as s e /m od °/ k
-PFC implemented in native Control Library
40
Tm as s e /proc e ss us ° /r
m in
30
Endothermicity Estimator and Feed-forward
0
20
40
60
80
1 20
1 40
Commande en pression à TEC constante
100
TEC
90
50
10 0
TSC
80
70
45
Jacket input temp
40
60
TMASSE
50
Internal temperature
40
35
30
30
Endothermicity estimator
20
PRESSION/10
10
25
0
20
0
1
2
3
4
5
6
7
8
9
4
x 10
YES BUT: …..!
ASSESSMENT ?
« Ok .. But my problem is more complex and very specific… »
- MATURE TECHNIQUE
Industrial resistance ?
- CONTROLERS and TECHNOLOGY : OK !
- lack of basic training in modelling
- no clear understanding of hierarchical control
- Validated by numerous world-wide applications in most
industrial domains
- a priori year budget and no « Pay Out Time » approach
- Short PAY OUT TIME…..
- but…
No conflict with PID : both techniques are used:
If PID works well (e.g, FIC) - use it
If not, use another tool…
85% of the work and 95% of the adrenalin
DYNAMIC FP MODELLING
Who is going to Implement
Predictive Control ?
Industrial Prospective
Since the main problem lies in Modelling
: The expert in Modelling ? ( Education ?)
- The local industrial Control Instrumentist ?
- The DCS/ PLC vendor ?
- An Engineering Company ?
- -------------------------------------------------------------
Integrated design:
To take into account the posibilities of
Advanced Control in the early design of the
process (e.g: CCV in Aircraft design)
: less steel, less concrete….
: more « brain juice »
Example in Chemical Engineering !
..since Control is made « simple », the Control expert has eliminated himself !
BATCH REACTOR: Enthalpic Control
2030€
1 Pump
1 Variator
2 Temp. sensors
Temperature source : ts
TI
1
TIT
3
TIT
2
Heating rod : P
SIC
2
Teaching control in Technical schools
-
39 Technical schools in France
≠ 660 Young technicians per year
PFC was taught to 23 profs of 4 academia
Lectures and laboratory works on PLCs
and laboratories processes.
- The first 125 young technicians went out
to industry in 2008, completely at ease with
PFC…
To be continued….
Thank you for your attention….
jacques. richalet @ wanadoo.fr
ACADEMIC CHALLENGE.!
There is a post-script…!
-----------------
A little more difficult ?
Time delay
Non minimum phase
Integrative
Unstable pole
Pure oscillator Time constant
Set point : ramp with no lag error
Constraints on MV and a state variable CV
H(s)= K.exp(-T1.s).(1-T2.s) / s.(1-T3.s).(s2+
ω
bon appétit…
2).(1+T4.s)