RTD software for identification of spatially localised models and data ...

daughterduckΠολεοδομικά Έργα

15 Νοε 2013 (πριν από 3 χρόνια και 8 μήνες)

83 εμφανίσεις

RTD software for
identification of spatially
localised models

and data standardisation

R.Žitný,
J.Thýn

Czech Technical University in Prague

.
History of RTD software
development at CTU


Radiotracers

group

UVVVR

since

1965
.



First

generation

of

RTD

(mainframe),

1973
-
1980


Second

generation

(HP

Basic),

1978
-
1989


Third

generation

(IBM

PC),

1989
-
1999


Fourth

generation

?

First generation of RTD


RTD1
Laguerre functions
RTD2
Time domain
RTD3
Identification
RTD interpret

Portability

(Fortran

4
,

file

oriented

I/O)


Transparency

(easy

modifications)


Simple

(user

friendly)

definition

of

batch


Example of batch

(identification)


INPUT(
1
)


INPUT(2)


NORM2(1,2)


ECOEF2(1,2,3)


TPOINT(3)


STOP



data (t
i
,c
i
) for time curve 1


inlet stream


data (t
i
,c
i
) for time curve 2


outlet stream.

Third generation of RTD

RTD0
c(T) processing
non-equidistant functions
corrections
RTD1
Time/Fourier domain
convolution/deconvolution
splines, Laguerre, FFT
RTD2
Numerical sol. ODE
identification
interactive models MDF
RTD


Interactive

Fortran
77
,

(Matrix

Editor,

PF)


Regularisation

in

time,

Fouriere,

Laguerre

domain


Identification

of

mixed

type

parameters

(integer/real)


Definition

of

models

using

MDF


Flow unit characterised by
(t,E or by Fourier series)
Convolution and deconvolution
IMPULSE RESPONSES
Flow unit characterised by
(first order, nonlinear)
Numerical integration
DIFFERENTIAL EQUATIONS
RTD1
RTD2
Black
-

Gray box analysis

Model definition file (MDF)

C______1 series & backmixing


[P1] Backmixing ratio f=@1F8.2


[P2] Mean residence time Te=@2F8.2


[P3] Number of units @3F6.0

\
\
INIT

real tm,f,aux integer i

f=p(1) neq=p(3) tm=p(2)/neq c(1)=1/tm

\
\
MODEL

dc(1)=(x+f*c(2)
-
(1+f)*c(1))/tm

i=1

while i<neq
-
1 do

begin i=i+1


aux=(1+f)*c(i
-
1)+f*c(i+1)
-
(1+2*f)*c(i)


dc(i)=aux/tm

end

dc(neq)=((1+f)*(c(neq
-
1)
-
c(neq)))/tm

y=c(neq)

\
\
PARAM

Models defined by users


variable

flow/volume,


axial

dispersion


collimation

characteristics,



heat

transfer


multiple

inlets/outlets,



heterogeneous

system


batch

systems


RTD0 application



Steam velocity.

Venturimeter calibration;(NZ)


Effective volume, holdup
. Waste stabilization pond (Ph; Ml), Holding tanks (Alumina
industry) (Au), Rotary kiln (Cz)


Parallel flows, bypass,channeling.

Tank with settler
-

pilot plant (K), Ethylalcohol
reactor (Cz), Precipitation tanks (Au), Holding tanks (Au)


Mixing characteristic, axial dispersion


Recirculation flowrate ratio


Separation effect,tracer balance.

Cement industry
-
Cyclone (K)

Moments, peaks
variance, area, Peclet
Corrections
background raise, tail, decay
Decomposition
regression (exponential, power)
Z-transformation
variable flow
Operations
RTD1 application

Crosscorrelation
time delay, PRBS, frequency char.
Splines
(linear/cubic)
FFT
(sin,cos)
Laguerre functions
(De)convolution - regularisation
E(t) identification, response prediction
Systems with recycles
identification, response predictions
Operations

Flow rate measurement.

Steam velocity measur.(NZ), flowrate measurement (K), incinerator
(NZ
), temperature disturbances (extremely slow flowrate) (Cz)


RTD functions: E(t), F(t),
8
(t).
Fluidized catalytic cracking (Au, Fi), Settling tank;
waste water treatment (Cz), Heat exchanger (In,K), Evaporator (In), Aniline reactor (In), Indirect rotary dryer (In),
Mercury removal unit (Th), Electron beam gas chamber (K), Direct ohmic heating (Cz)


Smoothing, disturbance attenuation.
Waste water treatment (Cz)


Transfer functions, Frequency characteristics
, Waste water treatment
(Cz), furnace for glass (Cz)


RTD2 application

Multiple inlets
responses, identification
Impulse responses
identification
Time domain models
Predictions, identification
Laplace transformation
Laplace domain models
Identification, response prediction

Pilot plants.

Mixed Tank (K); Extractor (Tl), Hygienization irradiator for waste water (batch system)


Waste water treatment.

Tank of activated sludge (TAS), air tubes (Ge), aeration turbines (Cz)
Gold system of activation (Cz), Sedimentation tanks (Cz), Equalization unit (Cz) Facultative oxidation Pond (Ph,Ml)


Heat exchangers.

Tubular heat exchanger (In),(K)


Evaporation, dehydratation.

Semi
-
Kestner (sugar industry) (In), Rotary dryer (In),
Dehydratation rotational furnace (pigment production) (Cz)


Reactors.

Production of aldehydes (Cz), Production of aniline (In), FCC crackinkg (1)Au, (2) Fi),


Alumina production

(Au). Precipitators; digestors; agglomeration, holding tanks


Electron beams gas chamber

(K). Chamber with baffles, chamber without baffles


Disintegration effect

Hammer mill (Cz), drum furnace (Cz)


Process modelling, LPM
and CFD


Lumped Parameter
(LPM)


Computer fluid dynamics

(CFD)


Combined Models

(CM)

0
10
20
30
40
50
60
70
80
90
100
Past
Present
Future
LPM
CFD
CM
EXP
?
tracer
EXPeriments

Why spatially localised
models?


Improved description and design of processes





Improved diagnostics


Residence time and temperature time T(t)


Thermal or irradiation treatment


Reaction is where?
I(z)
z
H
1
H
2
u(
r)
D
Spatially localised LPM


Series





Parallel

Constant concentration in the view field

Distance=
?

Distance=?

Non
-
uniform concentration in the view field

LPM & collimation algorithm





c
C
x
C
x
c
C
x
C
x
core
wall




11
1
12
2
21
1
22
2
dc
dx
C
C
dc
dx
C
C
core
wall




11
2
12
2
21
2
22
2
FCC
-
Fluidised Cracking

Core
-

Anulus model

(c
core
, c
wall
=?)

EXP & collimation algorithm





Collimated

Detector

Direct

Ohmic

Heating

CFD & collimation algorithm





Collimated

Detector

Direct

Ohmic

Heating

FEM
-
COSMOS/M

CV
-
FLUENT (600000 nodes, 600 MB results)

CFD/EXP collimation algorithm

R
d
R
d
h
z
z
1
2
2
2



;
(
)
e
hr
z

S
R
b
eR
R
b
eR
b



1
2
1
2
2
2
arcsin
arcsin
dJ
Sz
r
z
c
t
r
r
z
dz
dr
rd




4
2
2
3
2




(
)
(
,
cos
,
sin
,
)
/
b
R
R
e
R
R




1
2
4
1
2
2
2
2
1
2
2
2
2
(
)
Focused collimation algorithm

Focused
collimated
detector
4th generation RTD software


Project file


Script


Programs

Optimised
parameters
$MDF
name
\\icon

{bitmap
file reference)
<!
comment

keywords
INTEGER, REAL,
IF THEN
WHILE DO
BEGIN END

operators
+ - * / ** = <= >= <> | & ()

functions
sin,cos,exp,log,
abs,min,
max,
erf,
gama,
rnd,
atn,
bj,by (
Bessel
funct.,
Laguerre,
Legendre, Cheb.
polyn.)
COLC(), COLS() -
collimators

system
variables
T,X,Y,C(),DC(),P(),NEQ,NP,..
>
\\INITIAL
....
initial
conditions
\\MODEL
...
diff.
equations
defining model
$PAR
name
name MDF
<! model
parameters>
p
1
,...
$CT
name
<!time course (
dt=time step)>
\\dt
c
1
....
c
N
$CF
name
name CT
<! Fourier transform>
\\df
Re
1
Im
1
...
Re
N
,
Im
N
$JOURNAL
name
<! >
$MDF-FTP
name
<!
model definition >
URL
$COLLIM
name
<!
comment>
geometry
,window,...
$CONNECT
name
name XYZ
<!connectivity matrix for
CFD results>
el
1
M
1
i
1
...i
M1
...
$XYZ
name
<!
nodal points coordinates
(CFD results)>
coordinate system
1 x
1
y
1
z
1
...
$CTN
name
name XYZ
<! Predicted nodal values
at a specified time>
time
1 c
1
2 c
2
...
$TUPLEXIN
name
<!
example of "foreign"
CFD program>
Data manager
(
filters)
Model solver
(MDF)
numerical
integration
Model solver
(
fixed models)
numerical
integration
FFT
Deconvolution
(
regularisation)
R
xx

R
xy
Convolution
Normalisation
Comparion
(
optim. criteria)
Collimator
/detector
Fluent
conversion
Cosmos
conversion
SCRIPT
sequence of
called
programs and
specification
of data, i.e.
PROG [names
of sections $
..]
Working files
OPTIM
Conventions adopted:
-

section begins with
$
in the
first column
-

name
-arbitrary name of a
specific instance
-

<! ...> any comments
Conceptual

scheme

CORE


Project file


Script


Programs

FRONT
END


Windows


Visual studio?


Linux

GTK library

for

W
indows and
U
nix

UNIX or

Windows ?

B.G.makes
stars from
secretaries
and idiots
from experts