Diseño y Aplicaciones del PFR
Diseño y Aplicaciones del PFR
TURNO: B
I. INTRODUCCION
A medida que el líquido fluye por el PFR, se mezcla en dirección radial, pero la mezcla
no ocurre en la dirección del eje. Es decir, cada recarga de fluido hecha por el pistón se
considera una entidad separada que va bajando por la tubería. Sin embargo, el tiempo
transcurre a medida que el fluido (líquido o gas) se mueve corriente abajo. Así, hay una
dependencia implícita del tiempo, incluso en estado estacionario, en los problemas que
implican un PFR. No obstante, debido a que la velocidad del fluido (v) en el PFR es
constante, el tiempo y la distancia recorrida hacia abajo (x) son intercambiables y t = x/v.
Esto es, el fluido de un movimiento de pistón siempre toma un tiempo igual a x/v para
viajar la distancia x en el reactor.
• A medida que el liquido fluye por el PFR, se mezcla en dirección radial, pero la
mezcla no ocurre en la dirección del eje. Es decir, cada recarga de fluido hecha
por el pistón se considera una entidad separada que va bajando por la tubería.
Sin embargo, el tiempo transcurre a medida que el fluido (líquido o gas) se mueve
corriente abajo. Así́, hay una dependencia implícita del tiempo, incluso en estado
estacionario, en los problemas que implican un PFR
• Técnicamente no existe
• Alta velocidad
C. Condiciones
• Estado estable
• Densidad constante
El modelo del PFR considera que el fluido se desplaza como tapón, es decir, sin
gradientes radiales de concentración, temperatura o velocidad de reacción.
A medida que los materiales que reaccionan entran y fluyen axialmente por el reactor
son consumidos y la conversión aumenta en toda la longitud del mismo.
UNIVERSIDAD NACIONAL DE SAN AGUSTÍN
Facultad De Ingeniería De Proceso De La UNSA
ESCUELA DE INGENIERIA DE MATERIALES
D. Suposiciones:
• El Campo del flujo corresponde a un modelo de flujo tapón (PFR), que considera
la velocidad uniforme radialmente.
Mover la materia de un lugar a otro en el espacio, sin que haya alteración de sus
características internas.
PARTES
1 salida de agua,
2 reactor con lecho sólido,
3 depósito de aire,
4 entrada de reactivos A/B,
5 entrada de agua,
6 salida de producto,
7 manguito para sensor de
conductividad y temperatura
(incluido en la CE 310),
8 doble camisa para agua
El modelo de PFR considera que el flujo se desplaza como tapón, es decir, sin
gradientes radiales de concentración, temperatura o velocidad de reacción. A medida
que los materiales que reaccionan entran y fluyen axialmente por el reactor son
consumidos y la conversión aumenta en toda la longitud del mismo. Para desarrollar la
ecuación de diseño del PFR. Primero multiplicamos ambos lados de loa ecuación de
REACTOR DE
FLUJO PISTÓN
diseño de este reactor por -1. En seguida, expresamos la ecuación del balance molar
para la especie A en la reacción como sigue
−𝑑𝐹𝐴
= −𝑟𝐴
𝑑𝑉
𝐹𝐴 = 𝐹𝐴0 − 𝐹𝐴0 𝑋
Derivando
𝑑𝐹𝐴 = −𝐹𝐴0 𝑑𝑋
𝑑𝑋
𝐹𝐴0 = −𝑟𝐴
𝑑𝑉
𝑋
𝑑𝑋
𝑉 = 𝐹𝐴0 ∫
0 −𝑟𝐴
Para realizar las integraciones en las ecuaciones de diseño del reactor intermitente y
PFR, así como para evaluar la ecuación de diseño de CSTR, es necesario saber cómo
varia la velocidad de reacción −𝑟𝐴 con la concentración (y, por lo tanto, la conversión)
de la especie que reacciona.
● Operacion continua
UNIVERSIDAD NACIONAL DE SAN AGUSTÍN
Facultad De Ingeniería De Proceso De La UNSA
ESCUELA DE INGENIERIA DE MATERIALES
B. Desventajas
● Gradientes de temperatura
TIPO DE REACTOR
CARACTERISTICAS VENTAJAS DESVENTAJAS
FUENTE:YOUTUBE
FUENTE: YOUTUBE
V. APLICACIONES
Los PFR se utilizan en varias industrias, tale como:
Petróleo
UNIVERSIDAD NACIONAL DE SAN AGUSTÍN
Facultad De Ingeniería De Proceso De La UNSA
ESCUELA DE INGENIERIA DE MATERIALES
Petroquímica
Polímero
Farmaceútica
Tratamiento de Desechos
Especialidades Químicas
Energía Alternativa
Alimentos y bebidas
Carbonización
Deshidrogenación
Hidrogenación
Hidrocraqueo
Hidroformilación
Descomposición oxidativa
Oxidación Parcial
Polimerización
Reformación
El hogar
Ropa
Automóviles
Construcción
Electrónicos
REACTOR DE
FLUJO PISTÓN
2. INTRODUCCIÓN
En el reactor tubular, la reacción se inicia con el calor del calentamiento del agua.
La figura 01 ilustra un perfil de temperatura de un reactor tubular a alta presión y
UNIVERSIDAD NACIONAL DE SAN AGUSTÍN
Facultad De Ingeniería De Proceso De La UNSA
ESCUELA DE INGENIERIA DE MATERIALES
4. REACTOR TUBULAR
5. FOULING EN EL REACTOR
Esta separación de fases se produce en la región más fría del tubo del reactor
tubular. La figura representa el perfil de temperatura radial en el tubo del reactor
y la dirección del flujo de calor. Donde, Tw es la temperatura del tubo de la pared
UNIVERSIDAD NACIONAL DE SAN AGUSTÍN
Facultad De Ingeniería De Proceso De La UNSA
ESCUELA DE INGENIERIA DE MATERIALES
6. MECANISMO DE DEFOULING
7. CONCLUSIONES
A través de este estudio se puede ver que los resultados en los geles CAT1 están
directamente influenciados por la formación de incrustaciones en el reactor,
debido al aumento del tiempo de residencia en el reactor. Las condiciones físicas
y térmicas en el reactor y otros pasos del proceso de producción EVA que
favorecen una reducción del tiempo de residencia no son importantes para la
reticulación y la consiguiente generación de geles en el polímero. La temperatura
tiene un papel decisivo en la viscosidad de la mezcla en el medio de reacción,
confirmando el procesamiento del polímero en todo el reactor y reduciendo el
tiempo de residencia para el ensuciamiento.
1. Calcular:
a) La conversión que se obtiene en el reactor de mezcla completa
b) Volumen del reactor de flujo de pistón que nos proporcionaría
una conversión igual a la del apartado a).
B. Resolucion
a) Para l conversión que se obtiene en el reactor de mezcla
completa
𝑉 𝑋𝐴
=
𝐹𝐴𝑂 −𝑟𝐴
−𝑟𝐴 = 𝑘𝐶𝐴
𝐶𝐴 = 𝐶𝐴𝑜 (1 − 𝑋𝐴 )
−𝑟𝐴 = 𝑘𝐶𝐴𝑜 (1 − 𝑋𝐴 )
𝑉 𝑋𝐴
=
𝐹𝐴𝑂 𝑘𝐶𝐴𝑜 (1 − 𝑋𝐴 )
UNIVERSIDAD NACIONAL DE SAN AGUSTÍN
Facultad De Ingeniería De Proceso De La UNSA
ESCUELA DE INGENIERIA DE MATERIALES
CAo 𝑉 𝑋𝐴
=
𝐹𝐴𝑂 𝑘(1 − 𝑋𝐴 )
1𝑉 𝑋𝐴
=
2 0.7(1 − 𝑋𝐴 )
𝑋𝐴 = 0636
𝑋2
𝑉 𝑑𝑋𝐴
=∫
𝐹𝐴𝑂 0 −𝑟𝐴
𝑋2
CAo𝑉 𝑑𝑋𝐴
=∫
𝐹𝐴𝑂 0 𝑘(1 − 𝑋𝐴 )
1𝑉 −1 𝑋2 −𝑑𝑋𝐴
= ∫
2 𝑘 0 (1 − 𝑋𝐴 )
1𝑉 −1
= ln(1 − 𝑋𝐴 )x=0, x2= 0.636
2 𝑘
−2
𝑉= ln(1 − 0.636) = 2.9 𝐿
0.7
REACTOR DE
FLUJO PISTÓN
VIII. CONCLUSIONES
El reactor PFR es más eficiente que un reactor CSTR, porque para una misma
conversión el reactor PFR necesita menos volumen comparado al CSTR.
Estos reactores trabajan en estado estacionario. Es decir, las propiedades en
un punto determinado del reactor son constantes con el tiempo. En este tipo de
reactor la composición del fluido varia de un punto a otro a través de la dirección
del flujo, esto implica que el balance para un componente dado de la o las
reacciones químicas implicadas o debe realizarse en un elemento diferencial de
volumen
La acumulación en el reactor PFR, es 0 porque cuando es un reactor estable, ya
que los Moles de A en entrada en un lapso de tiempo serán iguales a los moles
de A que reaccionas + los moles de A de salida. Por eso la acumulación de A
será 0
Entre las aplicaciones más importantes y en las cuales ingeniería de materiales
puede tener un campo de trabajo sería, Petroquímica, tratamiento de residuos,
energías alternativas.
El reactor PFR entrega un producto de mayor calidad que los anteriores vistos
IX. BIBLIOGRAFÍA
FOGLER, Scott. Elementos de ingeniería de las reacciones químicas. Pearson.
Cuarta edición.
RÍOS, Italo. (2015). Determinación y simulación de variables termodinámicas en
reactores CSTR y PFR mediante Software HYSYS V8.0. Tesis. Universidad
Nacional de la Amazonía Peruana.
ALVAREZ, A., MYERSON, A. (2019). Caracterización del Mezclado en un
Reactor Tubular con Mezcladores Estáticos para la Cristalización Continua de
Productos Farmacéuticos. 7th Latin American and Caribbean Conference for
Engineering and Technology.
Pare Instrument Company. Reactores Tubulares de Flujo Continuo 5400.
Obtenido de [Link]
continuous-flow-tubular-reactors/
VELÁSQUEZ, Uriel. Reactor diferencial. Recopilado de:
[Link]
62121584
REACTO
R PFR
DEFINICIÓN
• El reactor tubular (PFR) se usa para modelar transformaciones químicas
de compuestos que se transportan en sistemas que parecen tuberías.
• A medida que el liquido fluye por el PFR, se mezcla en dirección radial,
pero la mezcla no ocurre en la dirección del eje. Es decir, cada recarga de
fluido hecha por el pistón se considera una entidad separada que va
bajando por la tubería. Sin embargo, el tiempo transcurre a medida que
el fluido (líquido o gas) se mueve corriente abajo. Así́, hay una
dependencia implícita del tiempo, incluso en estado estacionario, en los
problemas que implican un PFR
• Se dice de un sistema o proceso que está en estado estacionario si las
variables que definen su comportamiento , respecto del tiempo,
permanecen invariantes.
Consideraciones
• Modelo de flujo en tuberías
• Técnicamente no existe
• Velocidad del fluido es la misma en toda la sección trasversal de la
tubería
• No hay fricción en las pareces
• Alta velocidad
• Flujo turbulento (imposible en laminar)
Condiciones
• PLUGS infinitos (cilindros infinitos)
• Perfectamente mezclados y uniformes (radial)
• Gradiente de concentración/reacción (axial)
• Ecuación diferencial (por ser “plugs” infinita)
• Fluidos (líquidos, gas, slurries)
• Flujo turbulento (flujo pistón)
• Estado estable
• Densidad constante
PFR
• El modelo del PFR considera que el fluido se desplaza como tapón, es decir, sin
gradientes radiales de concentración, temperatura o velocidad de reacción. A medida
que los materiales que reaccionan entran y fluyen axialmente por el reactor son
consumidos y la conversión aumenta en toda la longitud del mismo.
PFR PLUG FLOW REACTOR
Reacciones en fase
Gaseosa
Suposiciones :
• El Campo del flujo corresponde a • La concentración de las especies varia
un modelo de flujo tapón (PFR), axialmente a lo largo del reactor.
que considera la velocidad uniforme Consecuentemente lo hace la velocidad
de reacción. La concentración no cambia
radialmente. en dirección radial para una
determinada posición axial.
CARACTERÍSTICAS
Se realiza una reacción con cambios en la concentración, la presión y la
temperatura, en la dirección axial.
• La conversión que se alcanza depende del volumen, el tiempo espacial, la
velocidad, el flujo y la concentración de alimento.
• El flujo de fluido a través es ordenado. Como consecuencia, el proceso de
transporte dominante es la advección.
Que es la advección?
Mover la materia de un lugar a otro en el espacio, sin que haya alteración
de sus características internas.
OTRAS CARACTERÍSTICAS
PARTES
PARTES DE UN REACTOR TUBULAR
FOTOCATALÍTICO (UV-A)
MODELOS
Fotoreactor tubular «Tube»
𝐹𝐴 = 𝐹𝐴0 − 𝐹𝐴0 𝑋
• Derivando
𝑑𝐹𝐴 = −𝐹𝐴0 𝑑𝑋
𝑑𝑋
𝐹𝐴0 = −𝑟𝐴
𝑑𝑉
• A continuación, separando las variables e integrándolas con las
fronteras 𝑉 = 0 cuando 𝑋 = 0 obtenemos el volumen de PFR
necesario para alcanzar una conversión especifica 𝑋:
𝑋
𝑑𝑋
𝑉 = 𝐹𝐴0 න
0 −𝑟𝐴
• Control automático de la
producción .
• El producto tiende ha ser
de mayo calidad en un
reactor.
• Su operación es continua.
DESVENTAJAS DE UN
REACTOR PFR
• Costo de adquisición elevado.
• No se recomienda para reacciones
de tiempo prolongado.
• Gradientes de temperatura.
• Difícil control de temperatura.
• Las paradas y limpiezas pueden
ser costosas.
VENTAJAS Y DESVENTAJAS
TIPO DE REACTOR CARACTERISTICAS VENTAJAS DESVENTAJAS
Es un reactor sin entrada ni salida solo Altas conversiones de reaccion con el Elevado costo de mano de obra.
BATCH tiene agitador que homogeneiza la incremento del tiempo.
mezcla. Producto variable
Simple construcción. en comparacion con los R.
continuos.
Trabaja con reaccciones en fase liquida Operación continua Baja conversión por unidad de
CSTR volumen
Mezcla perfecta simplicidad de construcción
Trabaja con reacciones en fase gaseosa Altas conversiones de reaccion por unidad Gradientes de temperatura
PFR de volumen
Velocidad radial uniforme Difícil control de temperatura
alta velocidad de reaccion
Concentracion uniforme en direcion axial buena transferencia de calor. Las paradas y limpiezas pueden ser
costosas
Recomendado para altas presiones
La regeneración del catalizador requiere del Buena distribucion de calor Este sistema presenta dificultades en el
PBR uso de gases. control de temperatura debido a la
formación de zonas calientes y frías en
SIMULACION
APLICACIONES
EN LAS INDUSTRIAS
PETRÓLEO FARMACEÚTICA
PETROQUÍMICA
TRATAMIENTO DE DESECHOS ESPECIALIDADES QUÍMICAS
HIDROGENACIÓN
TRATAMIENTO DE DESECHOS
TRATAMIENTO DE DESECHOS
Reactor Tubular de Flujo
Continuo 5400, Parr Instrument
Company
CARACTERIZACIÓN DEL MEZCLADO
EN UN REACTOR TUBULAR CON
MEZCLADORES ESTÁTICOS PARA LA
CRISTALIZACIÓN CONTINUA DE
PRODUCTOS FARMACÉUTICOS
OBJETIVO:
• Determinar experimentalmente el grado de dispersión o mezclado axial en un
sistema de flujo tubular equipado con mezcladores estáticos tipo Kenics a todo
lo largo del reactor.
Forma parte de la evaluación de un nuevo proceso
contínuo de cristalización de productos
farmacéuticos que permita producir cristales con
una variación muy reducida en su distribución de
tamaño de partícula
• Se ha explorado el uso de sistemas de flujo
tubular que permitan una operación cercana al
régimen ideal de flujo tapón, en la cual el
tiempo de residencia es constante, es decir, no
existe distribución de tiempo de residencia.
Figure 1 Temperature profile along the tubular reactor with two and three reactions
FOULING EN EL REACTOR
−𝑟𝐴 = 𝑘𝐶𝐴
𝐶𝐴 = 𝐶𝐴𝑜 (1 − 𝑋𝐴 )
−𝑟𝐴 = 𝑘𝐶𝐴𝑜 (1 − 𝑋𝐴 )
𝑉 𝑋𝐴
=
𝐹𝐴𝑂 𝑘𝐶𝐴𝑜 (1 − 𝑋𝐴 )
CAo 𝑉 𝑋𝐴
=
𝐹𝐴𝑂 𝑘(1 − 𝑋𝐴 )
1𝑉 𝑋𝐴
=
2 0.7(1 − 𝑋𝐴 )
𝑋𝐴 = 0636
𝑋2
𝑉 𝑑𝑋𝐴
=න
𝐹𝐴𝑂 0 −𝑟𝐴
𝑋2
CAo 𝑉 𝑑𝑋𝐴
=න
𝐹𝐴𝑂 0 𝑘(1 − 𝑋𝐴 )
1𝑉 −1 𝑋2 −𝑑𝑋𝐴
= න
2 𝑘 0 (1 − 𝑋𝐴 )
1𝑉 −1
= ln(1 − 𝑋𝐴 )x=0, x2= 0.636
2 𝑘
−2
𝑉= ln 1 − 0.636 = 2.9 𝐿
0.7
CONCLUSIONES
El reactor PFR es más eficiente
que un reactor CSTR
• Estos reactores trabajan en estado estacionario.
RESUMEN
El propósito de este trabajo es determinar experimentalmente el grado de dispersión o mezclado axial en un
sistema de flujo tubular equipado con mezcladores estáticos tipo Kenics a todo lo largo del reactor. El trabajo
forma parte de la evaluación de un nuevo proceso contínuo de cristalización de productos farmacéuticos que
permita producir cristales con una variación muy reducida en su distribución de tamaño de partícula. Los
resultados obtenidos en la investigación muestran que el sistema tiene un comportamiento cercano al flujo tapón
ideal, incluso a bajos valores de flujo volumétrico. El proceso de mezclado axial puede ser descrito
aceptablemente utilizando el modelo de dispersión propuesto en este estudio. Se encontró que existe una relación
lineal entre el flujo volumétrico del reactor tubular y el número de Peclet. Los parametros de dispersión obtenidos
serán utilizados para resolver un modelo más complejo de cristalización empleando balances de población.
Palabras claves: mezcladores estáticos, cristalización, productos farmacéuticos, dispersión, reactores tubulares
ABSTRACT
The purpose of this work is to determine experimentally the extent of dispersion or axial mixing in a tubular
system with Kenics type static mixers along the reactor. This work is part of a research aimed to asses a new
continuous crystallization process for pharmaceutical products intended to produce small crystals with a narrow
size distribution. Results obtained in this study show that the system approach an ideal plug flow system, even at
low volumetric flowrates. The axial mixing process can be described with the dispersion model proposed in this
study. A linear relationship between volumetric flowrate and Peclet number was found. Dispersion parameters
obtained in this work will be utilized to solve a complex model for crystallization using population balances.
Keywords: Static mixers, crystallization, pharmaceutical products, dispersion, plug flow reactors
1. INTRODUCCION
La cristalización es un proceso de separación y purificación ampliamente utilizado en la industria farmacéutica,
química y de alimentos (Myerson, 2002). Con el propósito de obtener cristales con una variación muy reducida en
su distribución de tamaño de partícula, se ha explorado el uso de sistemas de flujo tubular que permitan una
operación cercana al regimen ideal de flujo tapón, en la cual el tiempo de residencia es constante, es decir, no
existe distribución de tiempo de residencia (Van Wageningen, 2005).
2. METODO EXPERIMENTAL
APARATOS
Un sistema de flujo continuo fue construido y utilizado para los experimentos de caracterización del proceso de
mezclado. El sistema consiste en un reactor tubular de vidrio de 2.4 metros de longitud y 12.7 milímetros de
diámetro interno. El flujo a través del reactor es controlado con una bomba peristáltica (modelo Masterflex L/S
con velocidad variable, del proveedor Cole Parmer). En el interior del reactor se instalaron mezcladores estáticos
tipo Kenics, que consisten en una serie de hélices cortas de longitud fija (L=1.5D) rotados en un ángulo de 90°
entre si (Hobbs, 1997). La Figura 1 muestra un diagrama esquemático de los mezcladores estáticos.
1.5 D
D
Un espectrofotómetro infrarrojo (modelo IdentifyIR, del proveedor Smiths Detection Technology) con un rango
de longitud de onda de 4000 a 650 cm-1 fue utilizado para realizar mediciones de concentración de glicina en
solución acuosa.
MATERIALES
La glicina es el más simple de los aminoácidos que se encuentran comunmente en la mayor parte de las proteínas
de los seres vivos. En este estudio, se utiliza una solución acuosa de glicina como un trazador para caracterizar el
grado de mezclado alcanzado en un reactor tubular. La glicina utilizada en este estudio (grado reactivo, 98%) fue
PROCEDIMIENTO
a) Mediciones de concentración de glicina
La concentración normalizada de glicina en solución acuosa se utiliza para obtener experimentalmente la curva de
distribución de tiempo de residencia en el reactor tubular. Se desarrolló un método analítico para medir la
concentración de glicina en agua utilizando una técnica de espectroscopia infrarroja. El espectro infrarrojo de la
solución de glicina se muestra en la Figura 2, en donde se observa un pico característico en la longitud de onda de
1590 cm-1 (Doki, 2004).
0.10
0.08
0.06
absorbance
0.04
0.02
0.00
-0.02
Se preparó una curva de calibración utilizando soluciones de glicina de 10 a 200 mg/ml. Los datos de
concentración de glicina contra absorbancia a 1590 cm-1 siguen un comportamiento lineal (Ley de Beer-Lambert)
como se muestra en la Figura 3.
200
Concentración (mg/ml) y = 2149.2x - 0.9397
2
R = 0.9994
150
100
50
0
0 0.02 0.04 0.06 0.08 0.1
Absorbancia @ 1590 cm-1
∂ 2C ∂C ∂C
DL 2 − u x = (1) con la condición inicial C= Co a x<0 para t=0
∂x ∂x ∂t
con la condición inicial C= Co a x<0 para t=0 y C=0 a x>0 para t=0
y condiciones frontera
La solución con estas condiciones fronteras es complicada, pero se puede obtener una buena aproximación
utilizando las siguientes condiciones frontera
C = Co a x = - ∞ para t >= 0 C=0 a x = ∞ para t >= 0
La solución final es:
⎛C ⎞ 1⎡ ⎛1 1 (1 − θ / θ o ) ⎞⎟⎤
J (θ ) = ⎜ ⎟ = ⎢1 − erf ⎜ ⎥ (2)
⎝ Co ⎠ esc 2 ⎢⎣ ⎜2 Pe θ / θ o ⎟⎠⎥⎦
⎝
donde J(θ) es la función de distribución del tiempo de residencia, θ es el tiempo de residencia en el sistema, θo es
el tiempo medio de residencia (θo= volumen/flujo volumetrico) y Pe es el número de Peclet
ux L
Pe = (3)
DL
Cuando el número de Peclet se incremente y tiende a Pe = ∞, no existe difusión axial y el reactor se aproxima al
comportamiento de un sistema de flujo tapón, con tiempo de residencia constante como se muestra en la Figura 4.
Por otra parte, cuando el número de Peclet decrece y tiende a Pe = 0, la distribución de tiempo de residencia se
hace más amplia y el reactor se comporta como un tanque con agitación (Paul, 2004).
0.8
Pe = 500
Pe = 250
Pe = 167
0.6
Pe = 125
C/Co
Pe = 100
Pe = 20
0.4
Pe = 10
Pe = 2000
0.2 Pe = 10000
0
0 0.5 1 1.5 2 2.5
θ/θo
1.2
Funcion de distribucion de tiempo de
1
residencia J(θ/θo)
0.8
0.6
0.4
0.2
0
0 0.5 1 1.5 2 2.5 3 3.5 4
θ/θo
Reactor Flujo Mezclado Flujo Tapón Ideal
Flujo laminar Experimento @ 100 ml/min
El modelo matemático de dispersión descrito anteriormente en la sección 2.3 se utilizó para determinar el grado
de dispersión en el reactor tubular bajo distintas condiciones de flujo. Los datos experimentales se ajustan
aceptablemente al modelo, como se observa en la Figura 6. Los resultados muestran que al incrementar el flujo
volumétrico se incrementa el valor del número de Peclet. El número de Peclet está definido como la relación
entre el transporte por convección y el transporte por difusión. En este estudio, el número de Peclet se utiliza
como un indicador del grado de mezclado axial o dispersión. Mientras menor sea el valor del número de Peclet,
mayor será la dispersión en el sistema, y por otra parte un valor alto de número de Peclet indica un
comportamiento más cercano al flujo tapón ideal. Los altos valores de número de Peclet obtenidos
experimentalmente sugieren que el sistema tiene un comportamiento cercano al flujo tapón ideal.
1 1
0.8 0.8
C/Co
C/Co
100 ml/min 200 ml/min
0.6 0.6
Pe = 230 Pe = 290
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 0 0.5 1 1.5
θ/θo θ/θo
1.2 1.2
1 1
0.8 0.8
C/Co
0.4 0.4
0.2 0.2
0 0
0 0.5 1 1.5 0 0.5 1 1.5
θ/θo θ/θo
Al incrementarse el flujo, los valores de número de Peclet se incrementaron. Como se muestra en la Figura 6,
existe una relación lineal entre el flujo volumétrico y el número de Peclet. Estos resultados concuerdan con otros
valores reportados en la literatura (Raphael, 1995), (Brechtelsbauer, 2001), (Madhuranthakam, 2009).
300
Número de Peclet (Pe)
250
200
150
100
50
0
0 100 200 300 400 500
Flujo volumétrico (ml/min)
4. CONCLUSIONES
Los resultados obtenidos en la investigación muestran que el reactor tubular equipado con mezcladores estáticos
tipo Kenics tiene un comportamiento cercano al flujo tapón ideal, incluso a bajos valores de flujo volumétrico. El
proceso de mezclado axial puede ser descrito aceptablemente utilizando el modelo de dispersión propuesto en este
estudio. Se encontró que existe una relación lineal entre el flujo volumétrico del reactor tubular y el número de
Peclet. Mientras mayor sea el flujo volumétrico, el sistema se aproxima más al flujo tapón ideal. El presente
estudio es parte de una investigacion que busca desarrollar y modelar un proceso robusto de cristalización
continua Los parametros de dispersion (número de Peclet) obtenidos serán utilizados para resolver un modelo
más complejo de cristalización empleando balances de población.
.
REFERENCIAS
Brechtelsbauer, C., and F. Ricard (2001) “Reaction Engineering Evaluation and Utilization of Static Mixer
Technology for the Synthesis of Pharmaceuticals” Org. Proc. Res. Dev., Vol 5 (6), 646-651
Doki, N, H. Seki, K. Takano, H. Asatani, M. Yokota, and N. Kubota (2004) “Process Control of Seeded Batch
Cooling Crystallization of the Metastable r-Form Glycine Using an In-Situ ATR-FTIR Spectrometer and an
In-Situ FBRM Particle Counter” Crystal Growth and Design, 4, 949-953.
Hobbs, D.M., F.J. Muzzio. (1997) “Kenics static mixer: a three dimensional chaotic flow” Chemical Engineering
Journal, 67, 153-166
Autorización y Renuncia
Los autores authorizan a LACCEI para publicar el escrito en los procedimientos de la conferencia. LACCEI o los
editors no son responsables ni por el contenido ni por las implicaciones de lo que esta expresado en el escrito
Luiz Antonio Barbosa 1, Ademir Anildo Dreger 2, Ruth Marlene Campomanes Santana 3,
Eduardo Luis Schneider 4, Fernando Dal Pont Morisso 5,
1
Faculdades Integradas de Taquara, Av. Oscar Martins Rangel, 4500 (ERS 115), CEP 95600-000, Taquara, Rio Grande
do Sul , Brasil.
2
Faculdades Integradas de Taquara, Av. Oscar Martins Rangel, 4500 (ERS 115), CEP 95600-000, Taquara, Rio Grande
do Sul, Brasil.
3
Universidade Federal do Rio Grande do Sul - UFRGS, LAPOL - Laboratório de Materiais Poliméricos, Processamento
e Reciclagem Campus do Vale - Bloco IV - Prédio 43.426 - sala 115 Cx. Postal 15010 Av. Bento Gonçalves, 9500,
Agronomia, 91501-970, Porto Alegre, Rio Grande do Sul, Brasil.
4
Universidade Federal do Rio Grande do Sul - UFRGS, Departamento de Engenharia de Materiais, Campus do Vale,
Setor IV – Prédio 43426, Av. Bento Gonçalves, 9500, 91509-900, Porto Alegre, Rio Grande do Sul, Brasil.
5
Universidade Feevale, Laboratório de Estudos Avançados em Materiais, Curso de Mestrado Profissional em Tecnologia
de Materiais e Processos Industriais, Instituto de Ciências Exatas e Tecnológicas, Campus II, ERS 239, 2755, CEP
93525075, Novo Hamburgo, Rio Grande do Sul, Brasil.
e-mail: luizab@[Link], ademirdreger@[Link], [Link]@[Link], [Link]@[Link], moris-
so@[Link]
ABSTRACT
The fouling formation in the EVA production leads to an increase in the residence time of the material. This
result is a favorable condition to generation of the crosslinks during the polymerization. The crosslinking
allows raising the molar mass of the polymer, which after the polymerization reaction are characterized as
gels, thus generating problems in the product quality during production due to contamination of the resin.
This paper describes the fouling process in the reactor, the defouling mechanism, the crosslinking, the scis-
sion of chain and presents a study on the influence of the reaction temperature in relation to the formation of
gels by quantitative analysis by Optical Controller System - OCS. The average temperature rise of the reac-
tion at 10 ºC, helped to reduce the generation of gels due to the viscosity reduction over the reactor, which
reduces the polymer residence time in the walls of the tube and consequently the formation of fouling.
Keywords: Crosslinking, fouling, tubular reactor, EVA, OCS.
1. INTRODUCION
In plants producing poly(ethylene-co-vinyl acetate) EVA, a recurring problem is the formation of gels during
and after synthesis of polymerization. According to Henk (2015), the gels are any visible imperfection in
polyethylene film [1]. Contamination in polyolefin film by the formation of gels is a phenomenon that is dif-
ficult to predict, reproduce and resolve. This paper discussed the generation of gels formed from crosslinking
which is the result of unintended mixing of polymers of different density, molecular weight or type of
comonomer during polymerization.
According to Barbosa (2013), this defect can be characterized at the polyethylene film, due to the ap-
pearance of high molar mass molecules or other contaminants that do not melt, during processing of the
product [2]. This side effect, gels, is harmful to the production process, as it causes loss of product quality,
since its application is intended the fabrication of packaging for films what the correspond than half of con-
sumption. As Roman (1997). the film appearance quality is analyzed through the quantity of gels and the
importance this control is not only by the film appearance, but also because the infused, may weaken the ex
Autor Responsável: Luiz Antonio Barbosa Data de envio: 22/03/2017 Data de aceite: 03/01/2018
10.1590/S1517-707620180004.0565
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
truder balloon, sticking and causing stop in process of transformation EVA [3].
The objective of this work is to study the influence of the reaction temperature with respect to genera-
tion of crosslinking, with two and three reactions during the production of poly(ethylene-co-vinyl acetate)
EVA in a reactor tubular. The study considers that, in lower temperatures there is fouling formation in the
reactor tubes, which increases the residence time of the polymer and allows generation of crosslinks. The
importance of this study when compared to other production methods, lies in the increase of product quality
when applied to the manufacture of tubular films that requires high transparency.
300
Temperature (oC)
250
200
150
100
50
1900ral 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral
Tubular Reactor Length (dimensionless)
Profile Temperature 2 Reactions Profile Temperature 3 Reactions
Figure 1: Temperature profile along the tubular reactor with two and three reactions.
In the LDPE polymerization reactors also known as "high pressure low density" or HPLD, the
polymerization mechanism is a reaction of free radicals leading to the formation of long chain branches, LCB,
that can be as long as the main backbone polymer [15]. The free radical mechanism also leads to the for-
mation of short chain branches, typically one to five carbon atoms long [16].
Wall Tw
Tf Fouling
Flow
T1
Figure 2: Representation of the layer of fouling along the wall of the reactor tube [21].
At a given condition of temperature and reactor pressure, the temperature of the inner wall of the reac-
tor tube is the critical parameter that determines when fouling occurs, this must be controlled by coolant tem-
perature. This phase separation occurs in the coldest region of the tubular reactor tube. Figure 3 represents
the radial temperature profile in the reactor tube and the heat flow direction [21, 22]. Where, Tw is the reactor
wall tube temperature; Tf the fouling temperature; T1 the temperature at the fouling surface; and T P is the
process temperature.
1.7 Crosslinks
The crosslinks reactions are those leading to the formation of insoluble and infusible polymers in which the
chains are attached and form a three-dimensional network structure, called crosslinking [30,31, 32,33].
The reticule formation, hinders approximation of chains and the molecular interactions which reduces
the sliding of the same ones, increasing the mechanical strength and thus tend to become insoluble and infu-
sible polymer. During the crosslinking, branching of the polymer chains are recombined forming a polymer
with a molar mass very high or considered infinite. A polymer is considered to have infinite molar mass,
when the value measured is greater than the molar mass of the analysis equipment can measure [30,31,34].
According Tamboli et al, (2004) polyethylene crosslinking process occurs in four phases: initiation,
propagation, branching and termination. Figure 4 shows an example of the mechanism by crosslinking poly-
ethylene energy radiation abstraction of hydrogen and the generation of free radical in the polymer chain [31].
CH2 CH2 CH2 CH2 CH2 CH2 CH2 CH2+ CH2 CH2 CH2 CH2
Polyethylene
The free radicals formed propagate or recombine intramolecular or intermolecular form in the poly-
mer. In intramolecular recombination occurs cyclization of the polymeric matrix and at the intermolecular it
has the crosslinking. In the case of polyethylene, in an inert environment, prevails the crosslinking with an
increased molar mass due to degradation [24,35]. The simple crosslinking reaction of the EVA copolymer is
exemplified in Figure 5 through various polymer chains with functional groups capable of reacting together
to form chemical bonds C-C. If these polymer chains are exposed to certain conditions, wherein the function
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
al groups can react then these chains are connected by means of a new link C-C, forming a giant polymer
molecule in the medium reaction [36].
Figure 6 shows an image photograph of the gel taken at the OCS machine, of the EVA film, during
the film analizing process.
The results found were analyzed through statistical methods, to establish a correlation of results gels
found as two three polymerization reactions.
profile of temperature with two reactions was 11 °C above the temperature with three reactions.
1905ral 1900ral
1904ral 1900ral
1903ral 1900ral
1902ral 1900ral
1902ral 1900ral
1901ral 1900ral
1900ral 1900ral
1900ral 1900ral
1 11 21 31 41 51 61 71 81 91 101 111 121 131 141 151 161 171 181 191 201 211
Number of Samples
CAT1 - 3 reactions CAT1 - 2 reactions Average Temp Reactor 3 reactions Average Temp Reactor 2 reactions
Figure 7: Temperature profile along the tubular reactor with two and three reactions.
It can be seen in Figure 8, the temperature distribution along the tubular reactor at high pressure in a
steady state, where we can see and compare the profile with two and three reactions is during the experiment.
It was not possible to increase further the temperature according to the control limit in the fluidity index of
the resin.
230
210
Temperature (oC)
190
170
150
130
110
0 12 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral 1900ral
Figure 8: Temperature profile along the tubular reactor with two three reactions.
Figure 9, we can see the profile of velocity as a function of reaction temperature, concerning operating
time with two three reactions. The average results of speed for the better and the worst result with two reac-
tions the speed value was 10,91 m.s-1 with 223 gels and 10,91 m.s-1 with 866 gels. With three reactions with
speed of 10,88 m.s-1 and 185 gels, to 10,84 m.s-1 and 896 gels.
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
11,200
11,000
v (m/s)
10,800
10,600
10,400
10,200
0 10 20 30 40 50 60 70 80 90 100
Figure 9: Velocity profile of the gas along the tubular reactor with two three reactions.
The results analyses obtained through statistical methods using the MINITAB software, establish a
correlation between the amount of gels met in the two and three polymerization reactions.
This experiment was used as the upper limit of process control (UCL), the given index in the product
specification that is equal to 500 gels in CAT1.
Statistical analysis of the data regarding the film results in the polymerization process with two three
reactions performed through the graphical analysis of the two cases identifies that it is out of statistical pro-
cess control, signaling an alert in the production of the product, given that the chart points were outside the
control limits.
Figure 10 shows a chart of the results with control samples to the amount of gels during the produc-
tion of EVA with 3 reactions, which gave an average X equal to 421, the control lower limit value in the
(LCL) 137 and control upper limit (UCL) equal to 705 gels.
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
2500
2000
Nº Gels/m²
1500
1000
UCL=705
_
500
X=421
LCL=137
0
Figure 10: Control chart with the results gels in CAT1, with 3 reactions.
Figure 11 shows a control chart with results of the samples to the amount of gels during the produc-
tion of EVA 2 reactions, which gave an average X equal to 303, the control lower limit (LCL) 108, the
boundary UCL control top equal to 498 gels.
2500
2000
Nº Gels/m²
1500
1000
500 UCL=498
_
X=303
LCL=108
0
1 50 100 150
Samples
Figure 11: Control Letter to the results of gels in CAT1, with 2 reactions.
The normal analysis by Anderson-Darling was performed and demonstrated the data distribution was
not normal because P-value < 0.05. Because of robustness regarding the use of control charts and calculation
capability preferred to not transform the data set (using the Box-Cox procedure) in a lognormal, distribution
as shown in Figures 12 and 13. The data of the process reactions were normal in 3 with P-value < 0.05 and
not normal reactions were 2 with P-value < 0.044, very close to an acceptable value is 5%.
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
99 AD = 3,418
P-Value < 0,005
95
90
80
Percentage
70
60
50
40
30
20
10
5
0,1
100 1000
Rating Gels - CAT01
Figure 12: Test normal through the lognormal results in gels CAT1, with 3 reactions.
99 AD = 0,772
P-Value = 0,044
95
90
80
Percentage
70
60
50
40
30
20
10
5
0,1
100 1000
Rating Gels - CAT01
Figure 13: Test by normal lognormal results in gels CAT1, with 2 reactions.
Based on Six Sigma methodology, it can observe the process capability index in Figure 14 which
shows the results of the analysis of gels in the film through the OCS for 3 reactions. Through these indices
can be seen that the process is unable to meet specifications and produce products with a minimal margin
defects. The Cpk and Ppk are below the minimum specified values, and Cpk = 0.56 and Ppk = 0.10 (greater
than 1.67 are considered satisfactory). The ZBench index found for the total process capacity (Overall Capa-
billity), was ZBench = 0.63 Sigma, demonstrating that the process quality levels and product are not meeting
specifications. The number of defects per million in the process (Overall Performance) was 265349 PPM,
featuring 2 sigma process as not meeting the criteria for acceptance.
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
EVA 3 Reactions
Calculations Based on Lognormal Distribution Model
USL
P rocess Data O v erall C apability
LS L * [Link] 0,63
Target * [Link] L *
USL 500 Z.U S L 0,29
S ample M ean 420,762 P pk 0,10
S ample N 214
E xp. O v erall P erformance
Location 5,86367
P P M < LS L *
S cale 0,559758
P P M > U S L 265349
O bserv ed P erformance P P M Total 265349
P P M < LS L *
P P M > U S L 200935
P P M Total 200935
Capability index shown in Figure 15 through the lognormal distribution with the results of film analy-
sis by OCS to two reactions show that the process is able to meet specifications and produce product with a
minimum margin of defects. The Cpk and Ppk indices are below the minimum specified values, and Cpk =
0.95 and Ppk = 0.30 (greater than 1.67 to be satisfactory). The ZBench index found for the total process ca-
pacity (Overall Capabillity), was ZBench = 1.37, demonstrating that the levels of quality of process and
product are meeting the specifications. The number of defects per million in the process (Overall Perfor-
mance) was 84739.1 PPM, featuring The process currently meets the customer's requirements, and at 3 sigma.
EVA 2 Reactions
Calculations Based on Lognormal Distribution Model
USL
P rocess Data O v erall C apability
LS L * [Link] 1,37
Target * [Link] L *
USL 500 Z.U S L 0,91
S ample M ean 302,96 P pk 0,30
S ample N 177
E xp. O v erall P erformance
Location 5,61382
P P M < LS L *
S cale 0,43729
P P M > U S L 84739,1
O bserv ed P erformance P P M Total 84739,1
P P M < LS L *
P P M > U S L 67796,6
P P M Total 67796,6
4. CONCLUSIONS
Through this study it can be seen that the results in CAT1 gels are directly influenced by the formation of
fouling in the reactor, due to the increased residence time in the reactor. The physical and thermal conditions
in the reactor and other steps of the production process EVA which favor a reduction of residence time are
not important for crosslinking and consequent generation of gels in the polymer. The temperature has a deci-
sive role in the viscosity of the mixture in the reaction medium, confirming to the processing of the polymer
throughout the reactor and reducing the residence time for fouling.
Performance results in gels, are strongly related to the operating temperatures than the number of reac-
tions, but working with two reactions allowed raise the average temperature of the reaction medium and,
maintain control of the variables that determine the melt index.
This study may be expanded to other reaction conditions or process so that the statistical analysis of
the data are within the statistical process control.
5. ACKNOWLEDGMENTS
The authors thank PPGE3M-UFRGS and the University Feevale, for providing all infrastructure for the de-
velopment of this research.
6. BIBLIOGRAPHY
[1] HENK L. Gel formation during extrusion of LDPE and LLDPE., 2015 Retrieved from
[Link]
[2] BARBOSA, L. A. Estudo da geração de crosslinking durante a fabricação de EVA/PEBD. Dissertação
de [Link]., Universidade Feevale, Novo Hamburgo, Rio Grande do Sul, Brasil, 2013.
[3] ROMAN, A. Polietileno PEBD – Processos de Transformação. Editora Érica Ltda, São Paulo, 1997.
[4] KIPARISSIDES, C., VERROS, G., MACGREGOR, J. F. “Mathematical modeling, optimization, and
quality control of high-pressure ethylene polymerization reactors”, Journal of Macromolecular Science, v. 4,
pp.437-527, 1993.
[5] KWAG, B. G., CHOI, K. Y. “Modeling of a multistage high-pressure ethylene polymerization reactor”,
Chemical Engineering Science, v.49, pp.4959-4969, 1994.
[6] PEACOCK, A. J. Handbook of polyethylene, strutures, porperties, and aplications. Editado por Marcel
Dekker, Inc. New York, 2000.
[7] NEUMANN, G. A. Modelagem de um reator tubular de alta pressão para produção de PEBD. Disserta-
ção de [Link]., Universidade Federal do Rio Grande do Sul, Porto Alegre, 2001.
[8] CHIEN, I-L., KAN, T. W., CHEN, B. “Dynamic simulation and operation of a high pressure ethylene-
vinyl acetate (EVA) copolymerization autoclave reactor”, Computers & Chemical Engineering, v.31, n.3,
pp.233–245, 2007.
[9] ASTEASUAIN, M., TONELLI, S. M., BRANDOLIN, A., et al., “Dynamic simulation and optimisation
of tubular polymerization reactors in gPROMS”, Computers and Chemical Engineering, v.25, n. 4–6), pp.
509–515, 2001.
[10] GRILLER, D., INGOLD, K. U. “Persistent Carbon-Centered Radicals”, Accounts of Chemical Research,
v.9, n.1, pp.13-19, 1976.
[11] TOBITA, H. “Free-Radical Polymerization with Long-Chain Branching and Scission: Markovian Solu-
tion of the Weight-Average Molecular Weight”, Macromolecular Theory and Simulations, v.23, pp. 477–489,
2014.
[12] ASPEAK. Initiators, free-radical. Encyclopedia of Polymer Science and Technology, 2015. Retrieved
from [Link]
Free-Radical%20([Link]).pdf.
[13] PLADIS, P., KIPARISSIDES, C. “A comprehensive model for the calculation of molecular weight
long-chain branching distribution in free-radical polymerizations”, Chemical Engineering Science, v.53, n.18,
pp.3315-3333, 1998.
[14] KOLHAPURE, N. H. Computational fluid dynamics for design and optimization of tubular low-density
polyethylene reactors, Dissertation Doctor of Philosophy Major: Chemical Engineering, Iowa State Universi-
ty, 2001.
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
[15] FLORY, P. J. “The Mechanism of Vinyl Polymerizations”, Journal of The American Chemical Society,
v. 59, n.2, pp. 241–253, 1937.
[16] ROEDEL, M. J. “The Molecular Structure of Polyethylene I. Chain Branching in Polyethylene during
Polymerization”, Journal of The American Chemical Society, v.75, n. 24, pp.6110–6112, 1953.
[17] ERDEGHEM, V., PETER, M. M, LOGIST, F., et al., “Model based optimisation of tubular reactors for
LDPE production”, In: Preprints of the 8th IFAC Symposium on Advanced Control of Chemical Processes
The International Federation of Automatic Control Furama Riverfront, Singapore, 2012.
[18] KIPARISSIDES, C., VERROS, G., PERTSINIDIS, A. “On-Line Parameter Estimation in a High-
pressure Low-Density Polyethylene Tubular Reactor”, AIChe Journal, v. 42, n.2, pp. 440–454, 1996.
[19] MAN, C. W. Modelagem Matemática e Aplicação de Rede Neuronal no Sistema de Copolimerização de
Oleifinas a Alta Pressão, Tese de [Link]., Escola Politécnica, Universidade de São Paulo, 1995.
[20] SEBASTIAN, F., CASTAÑEDA-ZUÑIGA, D. M., NEUTEBOOM, J. D. P., et al., “Fouling in the High
Pressure LDPE Process: Experimental and Computational Investigation Approach”, Macromolecular Symo-
sia, v.360, pp. 78–86, 2016
[21] BUCHELLI, A., CALL, M. L., BROWN, A. L., et al., “Modeling Fouling Effects in LDPE Tubular
Polymerization Reactors. 1. Fouling Thickness Determination”, Industrial & Engineering Chemestry Re-
search, v. 44, pp. 1474-1479, 2005.
[22] MYKHAYLO, K., KUNDE, C., MANGOLD, M. “Simulation and Nonlinear Analysis of the Stagnant
Polymer Layer in a LDPE Tubular Reactor”, In: Proceedings of the 22nd European Symposium on Computer
Aided Process Engineering, pp. 17 - 20, London. Retrieved from, 2012.
[23] LACUNZA, M. H., UGRIN, P. E., BRANDOLIN, A., et al., “Heat Transfer Coefficient in a High-
Pressure Tubular Reactor for Ethylene Polymerization”, Polymer Engineering & Science, v. 38, n.6, pp. 992-
1013, 1998
[24] PAOLI, M. A. Degradação e Estabilização de Polímeros. Segunda versão on-line Retrieved from , 2008.
[Link]
[25] GOMES, L.B. Avaliação da degradação em solo simulado de filmes de polietileno com aditivos pró-
degradantes após exposição ao envelhecimento acelerado., Trabalho de Conclusão do Curso, Universidade
de Caxias do Sul, Caxias do Sul, 2010.
[26] CANEVAROLO Jr, S. V. Ciência dos Polímeros. Artiliber Editora Ltda, São Carlos, 2004.
[27] BEYLER, C. L., HIRSCHLER, M. M. Thermal decomposition of polymers. SFPE handbook of fire pro-
tection engineering, v. 2, pp. 111-131, 2002
[28] RODRÍGUEZ-VÁZQUEZ, M., LIAUW, C. M., ALLEN, N. S., et al., “Degradation and stabilisation of
poly (ethylene-stat-vinyl acetate): 1–Spectroscopic and rheological examination of thermal and thermo-
oxidative degradation mechanisms”, Polymer Degradation and stability, v.91, n.1, pp.154-164, 2006.
[29] SCOTT, G. “Initiation processes in polymer degradation”, Polymer Degradation and Stability, v. 48, pp.
315-324, 1995.
[30] KRUPA, I., LUYT, A. S. “Thermal properties of uncross-linked and cross-linked LLDPE/wax blends”,
Polymer Degradation and Stability, v.70, n.1, pp.111–117, 2000.
[31] TAMBOLI, S. M, MHASKE, S. T, KALE, D. D. “Crosslinked polyethylene”, Indian Journal of Chemi-
cal Technology, v.11, pp.853-864, 2004.
[32] LABARTA, J. A. R., OLAYA M. M., MARCILLA, A. “DSC and TGA study of the transitions involved
in the thermal treatment of binary mixtures of PE and EVA copolymer with a crosslinking agent”, Polymer, v.
47, n.24, pp.8194–8202, 2006.
[33] AKCELRUD, L. Fundamentos da Ciência dos Polímeros. Editora Manole Ltda, São Paulo, 2007.
[34] SOARES, B. G., LUCAS, E. F., MONTEIRO, E. E. C. Caracterização de Polímeros – Determinação de
peso molecular e análise térmica. E-papers Serviços Editoriais Ltda, Rio de Janeiro, 2001.
[35]. LUGÃO, A. B. Estudos da síntese por irradiação, da estrutura e do mecanismo de formação de poli-
propileno com alta resistência do fundido., Tese de [Link]., Instituto de Pesquisas Energéticas e Nucleares,
São Paulo, 2004.
BARBOSA, L.A.; DREGER, A.; SANTANA, R.M.C., et al. revista Matéria, v.23, n.4, 2018.
[36] STELESCU, M. D., MANAILA, E., CRACIUN, G., et al., “Crosslinking and grafting ethylene vinyl
acetate copolymer with accelerated electrons in the presence of polyfunctional monomers”, Polymer Bulletin,
v. 68, n.1, pp. 263-285, 2012
[37] CALLISTER, W. D. J. Fundamentos da Ciência e Engenharia de Materiais – Uma Abordagem Inte-
grada. LTC – Livros Técnicos e Científicos Editora S.A., Rio de Janeiro, 2006.
[38] RIBEIRO, R. P. Efeito da radiação gama no comportamento in vitro de um copolímero bioabsorvível.
Tese de [Link]. Ciências dos Materiais, Instituto Militar de Engenharia, Rio de Janeiro, 2006.
ROBUST GENERIC MODEL FEEDBACK UNDER MODEL UNCERTAINTIES:
APPLICATION OF A TUBULAR REACTOR FOR THE TREATMENT OF
INDUSTRIAL PLATING WASTEWATER
Ricardo Aguilar-López, Ricardo Acevedo-Gómez, María Isabel Neria González* and Alma Rosa
Domínguez-Bocanegra
Departamento de Biotecnología y Bioingeniería
CINVESTAV-IPN
Av. Instituto Politécnico Nacional No. 2508, San Pedro Zacatenco, México, D.F. 07360, MEXICO
E-mail.- raguilar@[Link]
Phone.- + 52 55 5747 3800, ext. 4307
*División de Ingeniería Química y Bioquímica
Tecnológico de Estudios Superiores de Ecatepec
Av. Tecnológico s/n Esq. Av. Carlos Hank González Col. Valle de Anáhuac C.P. 55210, Ecatepec Estado de México
ABSTRACT
The goal of this work is to describe the design of a feedback controller to regulate the Hexavalent
chromium (Cr (VI)) concentration in industrial plating wastewater in a tubular reactor. Cr(VI) is a major
pollutant present in industrial wastewaters common to the metal and mineral processing as well as to
plating industries. For control purposes, a dynamic model of a plug flow reactor with dispersion is
developed in the form of partial differential equations (PDE) via applying classical conservation
principles; this model is experimentally corroborated using laboratory data. The proposed controller is
based on uncertainty observer to infer unknown terms as diffusive transport and Cr(VI) kinetic rate; also
Generic Model Controller (GMC) is coupled with this estimation methodology to provide robustness
against model uncertainties. The controller leads to nonlinear PID plus double integral action, where the
tuning rules are given in terms of the observer and GMC gains. A theoretical frame is given in order to
show the stability properties of the closed-loop system under the proposed controller. Numerical
simulation illustrates performance of the methodology proposed; a comparison with observer based I/O
linearizing controller and a well tuned PID controller is done, which allows concluding that the observer
based Generic Model Controller exhibits better performance.
KEY WORDS: Cr(VI) treatment, GMC, Tubular reactor, Infinite dimensional models, Robust performance.
RESUMEN
El objetivo del presente trabajo es describir el diseño de un controlador retroalimentado para regular la
concentración de Cromo Hexavalente (Cr(VI)) de agua residual de la industria de galvanoplastía en un
reactor tubular. El Cr(VI) es uno de los más tóxicos contaminantes de esta industria. Para fines de
control, se diseña un modelo dinámico del proceso, el cual es un reactor flujo pistón con dispersión
descrito por ecuaciones diferenciales parciales que se basa en la aplicación de los principios básicos de
balances de masa; este modelo se corrobora experimentalmente con datos de laboratorio. El
controlador propuesto se basa en un estimador de incertidumbres con el fin de inferir los términos
relacionados con el transporte difusivo de masa y la cinética de remoción de Cr(VI), que se asumen
desconocidas; posteriormente, este estimador se acopla con un controlador de modelo genérico para
generar un esquema de control robusto contra incertidumbres. El diseño del controlador propuesto es
equivalente a una estructura de tipo PID más una acción doble integral no lineal, donde las reglas de
sintonizado se dan en términos de las ganancias del observador de incertidumbres y del propio
controlador genérico. Se proporciona el marco teórico que muestra que la metodología propuesta
produce estabilidad a lazo cerrado. Simulaciones numéricas muestran el desempeño de la metodología
184
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
Palabras clave: Tratamiento de Cr (VI), control de modelo genérico, reactor tubular, modelos de
dimensión infinita, desempeño robusto
1. INTRODUCTION
Tubular reactors are very important processes equipments in chemical industry because of their ability to
provide large continuous production, avoiding mechanical issues because of mixing with important
energy saving. Currently, the adequate operation and control of tubular reactors is a challenging issue
for control engineers because these systems exhibited distributed parameters; i.e. the reactor’s variables
depend on position and time. Tubular reactors are mathematically modeled by nonlinear Partial
Differential Equations (PDE) derived from first principles application (mass, energy momentum balances,
chemical kinetics, thermodynamics), reason why the analysis from the classical control theory point of
view (where the models are described by Ordinary Differential Equations) turns not to be enough.
In contrast, process engineers have been employing linear controllers, including PI/PID compensators.
There are many reasons for this, among them the adequate understanding of the process operation
given by the long history in the employment of linear compensators, besides others advantages such as
emergency manual modes, automatic and manual switching and set point tracking. However, despite
successful industrial operation of linear control, so far, there is a lack of theoretical results to back up its
regulation capacity and robust designs.
The problem of feedback control design for PDE has been considered by several authors (e.g. Balas,
1979; Chistofides, 1998; Chistofides and Dautidis, 1998; Froment and Bischoff, 1990), the resulting
controllers depend on spatial coordinates. The control designs mentioned above do not look satisfactory
at all because in industrial process the controllers act at the boundary of the process; where the control
output is the fluid flow at the inlet of the process (Froment and Bischoff, 1990; Aguilar et al., 2002).
Meanwhile, several papers describing alternative control designs for tubular reactors have been
presented. In these works, the main idea is to transform the PDE models (infinite dimensional model)
into finite set of ordinary differential equations (ODE) (finite dimensional models) via discretization
techniques such as finite differences, finite element, orthogonal collocation, etc. where the goal is to
regulate the process at some specific point, generally at process output conditions (Aguilar et al., 2002;
Dochain, Babary and Tali-mammar, 1992; González et al., 1999). However, these approaches lack
information because of considering the control design procedure on finite dimensional models;
consequently, the robustness of these methodologies is only locally assured.
In this work, a strategy that considers the distributed nature of tubular reactors is developed to design
the control avoiding information loss. An electrochemical tubular reactor for the treatment of Cr(VI) from
industrial plating wastewater is considered as study case, considering the diffusive transport and the
kinetic term as model uncertainties. An uncertainty estimator is coupled with the Generic Model Control
Law in order to provide proper robust properties, given by the integral action considered in this control
technique (Lee and Sullivan, 1988; Signal and Lee, 1992). The methodology proposed is equivalent to a
nonlinear PID plus double integral controller, it is compared to an observer based I/O linearizing
controller and a well tuned linear PID controller; computer simulation shows that the best performance of
the three methodologies is exhibited by the observer-based GMC controller
185
Journal of Applied Research and Technology
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
2. PROCESS DESCRIPTION
Cr(VI) appears to be carcinogenic; even in small quantities it irritates plant and animal tissues being a
strong oxidizing agent easily absorbed by the skin. Health effects of chromium are related to the valence
state of the metal and the time of exposure. Trivalent chromium (Cr [III]) is an essential dietary mineral in
low doses. Cr(VI) is generally considered 1,000 times more toxic than Cr (III) (Department of Health and
Human Services, 2000). Some electroplating, leather tanning, and textile industries release relatively
large amounts of chromium in water bodies; in addition, solid wastes from chromate-processing facilities
have been disposed improperly in landfills and they are sources of contamination for groundwater. This
water pollutant is able to display considerable diffusion rates through soils and aquatic environments.
Several treatment processes have been proposed in order to diminish Cr (VI) concentrations from
industrial wastes including chemical precipitation, ion exchange, evaporation and reverse osmosis. In
these kinds of methods, an excess dosage of 2.5 times the theoretical addition of chemical agents
(ferrous sulphate) must be applied in order to complete the reaction. Therefore, it produces big quantities
of sludge, increasing the cost of the process due to the treatment, handling and disposal of this sludge.
Electrochemical Cr (VI) reduction is an alternative process which has been studied and applied
successfully to remove Cr(VI) from wastewaters at laboratory level (Rodríguez et al., 2003). Reactions
(I), (II), (III) and (IV) are considered to be present during the electrochemical process as shown in Figure
1. In fact, electrochemical processes have been proposed as an attractive alternative for Cr (VI)
treatment because of their low process cost in comparison to traditional methods. After the electrolytic
treatment, water can be reused in rinsing baths or can be discharged complying with environmental
standards.
bulk liquid
Cr 3+
+ II -
Cr 6+
Fe (s)
3Fe 2+ 3Fe 3+
I
Cr3+ 3Fe 2+
anode III cathode
Cr 6+
Cr 6+
IV 3e -
Cr 3+
Several types of electrochemical reactors, batch and continuous stirred tank reactors, have been
previously studied (Rodríguez et al., 2004); however, mechanical mixing devices rise operating costs for
stirred reactors. Therefore, a tubular electrochemical reactor seems adequate to perform economical
and efficient removal of Cr (VI) from waste water.
186
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
In order to evaluate chemical kinetics, experimental data were obtained in a batch reactor of 170 L. The
electrochemical reactions that take place within the reactor can be summarized as follows:
In the anode:
Fe0 - 2 e- ⎯→ Fe 2+ (1)
In the solution:
The Cr (VI) reduction as function of time is described by a kinetic model obtained that exhibits goodness
of fit (Chi-square test) at 0.01 significance level (a 99% confidence level). The experimental data were
well fitted by a zero order equation at high Cr(VI) concentrations and by a first order reaction at low
concentrations. The overall rate equation (rCr) with respect to the Cr(VI) in the liquid (Equation 4), was
obtained employing the standard integral method (Froment and Bischoff, 1990). Experimental data and
model (solid line) predictions are shown in Figure 2. As it can be seen, the reaction time to reach 0.5 mg
of Cr(VI)/L in the wastewater was 55 minutes (note the large operation time). The value of 0.5 mg Cr
(VI)/L was chosen because it is the maximum concentration of Cr(VI) permitted by Mexican
environmental regulations.
187
Journal of Applied Research and Technology
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
- - - reaction I
o experimental
____ model
Time [minutes]
Figure 2. Cr (VI) concentration as a function of time at a I = 190 A/m2 in the reactor (Volume = 170
Liters)
In this work, an electrochemical tubular reactor containing a steel electrode is proposed to treat, in a
continuous way, plating wastewater containing hexavalent chromium. Metal dissolution was
electrochemically induced by applying a current density of 50 A·m-2 by means of a direct current power
supply. This process was performed in a plug flow reactor provided with steel electrodes, which
consisted of 62 plates where 31 worked as anodes and the remainder as cathodes. The alternating stack
of electrodes within the reactor was contained in an acrylic column of 1.295 m height and 0.064 m in
diameter, whereby a cathode followed each anode. The area of each steel plate was 0.0001625 m2; so,
total “sacrificial” electrode surface Se was 0.10075 [Link] column reactor (Figure 3) has four sampling
ports at equal distances in which samples were collected to be analyzed in order to observe changes in
Cr (VI) concentration as function of elapsed time and reactor distance. Conditions such as pH (adding
H2SO4) and current density were maintained at the same value throughout the processing time.
Synthetic wastewater solutions containing Cr(VI) were prepared using potassium dichromate (reagent
grade) with known concentration of Cr(VI) of 130 mg·L-1 in distilled water in order to reach pH=2. Cr (VI)
concentration was also measured by treatment with 1-5-diphenylcarbohidrazide method (APHA, AWWA,
WPCF, 1995); in this case, a Hach 3000 spectrophotometer was used and the absorption readings were
obtained at 540 nm.
188
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
V A
-
Effluent Cr(III)
4
1.15 m
3
Sampling ports
2 Electrode
arrangement
1
Peristaltic pump
Influent Cr(VI)
0. 38 m
Cr(VI) reservoir
0.30 m
The reactor was continuously feed with the aqueous solution containing 130 mg·L-1 of Cr(VI). It becomes
straightforward that a continuous Cr(VI) reduction of 98.5 % (1,5 mg/l) is achieved with a maximum flow
rate of 338.61 mL min-1. It is important to note that despite the high conversion reached, the Cr (VI)
concentration is 200% up of the Cr(VI) concentration required by the legislation (0.5 mg/l). A current
density of 50 A·m-2 was imposed after the electrochemical reactor was filled with the solution using a
tracer.
D
Dispersion coefficient was calculated as is it usual (Bird et al., 1962), resulting = 0.078 , used in the
uL
model of this work. This set of process conditions was previously reported by Soto et al. (2002).
One of the main drawbacks of the continuous operation for Cr(VI) treatment in wastewaters employing
electrochemical reactors is the large open loop residence time of the process needed to reach the
required Cr(VI) concentration at reactor output, as it is reported by Rodríguez et al. (2004a) and
Rodríguez et al., (2004b). The large residence time is related to low processed flow rate of wastewater,
which is undesirable for the reactor operating performance. In order to avoid it, some actions can be
considered. One of them, it is to increase the pH inside the reactor to improve the kinetic rate conditions,
but this action leads to high corrosion levels on the reactor. Another action is related to the increase of
the density of the electrical current applied to the reactor, raising the economic operation cost, being
inadequate too.
189
Journal of Applied Research and Technology
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
Applying differential mass balance for the Cr(VI) concentration in the fluid phase, considering dispersion
term in one-dimensional transport approach, the following dynamic plug flow reactor model is obtained:
Boundary conditions:
CCr = CCr 0 at z = 0 ∀t
∂ CCr
= 0 at z = L ∀t
∂z
Initial condition:
CCr = CCr * for t = 0 ∀ z
Momentum balance to describe the velocity field for the axial coordinate of the reactor, considering the
case where the momentum balance equation has not fluid dispersion:
∂ vz ∂v
= −vz z (6)
∂t ∂z
This model was solved via finite differences, considering the operating conditions described; model and
its simulation reproduces experimental data with satisfactory agreement (Figure 4).
Because of open-loop characteristics, operating policy for electrochemical reactors requires the
application of control procedures, in this way, the reactor can be forced (via feedback control) to reach
the Cr(VI) required by environmental legislation, to keep adequate processed wastewater flows and to
save energy. For regulation purposes, the PDE model developed for the reactor is employed to design a
robust controller. The considered measured process output is Cr(VI) concentration at reactor output
conditions and the corresponding manipulated input is the input flow rate (velocity boundary condition at
z=0) as usual.
In order to implement the control law, let us consider the following alternative Cr(VI) mass balance
representation, assuming incompressible liquid; therefore, there is uniform velocity profile for the flow of
disturbances along the reactor, i.e. vz ( z , t ) = u (t ) :
190
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
∂CCr ΔC
= ϕ ( z , t ) − u (t ) Cr (7)
∂t Δz
∂ 2CCr ∂v
ϕ (z, t ) = D
kC
Here − CCr z − 1 Cr + DAE (8)
∂z 2
∂ t 1 + k2CCr
DAE is the derivative approximation error of the Cr(VI) concentration gradient (i.e.
∂ CCr ΔCCr
= + DAE ) and u is the control input.
∂z Δz
ϕ is considered an uncertain term on the mass balance equation, it exhibits its own unknown dynamic
behavior as
∂ϕ
= γ (z, t ) (9)
∂t
In order to ensure that concentration of Cr(VI) will follow the desired behavior at reactor outlet, Generic
Model Control (GMC) law is proposed. GMC has proved to be an effective tool to regulate nonlinear
systems; it belongs to the family of generalized linearizing controllers, which under the assumption of
knowledge of perfect model for the system, this controller cancel nonlinearities, imposing desired
behavior. In particular GMC contains integral action that, as it is well known, induces robustness to the
closed-loop behavior of the system.
∂ CCr
t
= g1−1 (CCr − CCr − SP ) + ∫ g 2−1 (CCr − CCr − SP ) dt (10)
∂t 0
Equation 10 represents stable behavior; from Equations (9) and (10) and applying the GMC stands, the
ideal control law u can be obtained, as follows:
Δz ⎡ −1 ⎤
t
In order to ensure the corresponding restriction, ΔCCr is considered the difference between the outlet and
input Cr (VI) concentrations of the reactor, with Δz = L.
Note that the above controller is realizable if terms, especially ϕ, are known. However, ϕ is considered
uncertain, therefore standard GMC cannot be implemented. In order to supply this information, there are
methodologies related to observers. Aguilar-López and collaborators have tried several observer designs
for uncertain systems such as proportional, proportional-integral, sliding-mode, applied to systems
modeled by ODEs (Aguilar-López and Maya-Yescas, 2005; Aguilar-López, 2003).
The system given by Equations (7)-(9) cannot be seen as a standard observer structure, which consists
of a copy of the system plus correction of the measurement error; it is not possible since the term γ ( z, t )
is an unknown function of the system output CCr . Let us propose the following uncertainty observer:
191
Journal of Applied Research and Technology
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
• • •
ϕˆ = τ 1 (ϕ − ϕˆ ) + τ 2 (ϕ − ϕˆ ) (12)
This uncertainty estimator is a reduced order observer, which infers the uncertain term ϕ. Note that the
observer contains proportional and derivative actions, the aim of the derivative term is to improve the
speed of the estimation algorithm because it enhances anticipatory and stabilizing effects of derivative
actions and it is able to diminish the peaking phenomenon present in controllers based on observers
(Sussman and Kokotovic, 1991). Introducing this observer, the non ideal controller is given by
Δz ⎡ −1 t
⎤
u= (
⎢ 1 Cr
g C − CCr − SP ) + ∫ g 2−1 (CCr − CCr − SP ) − ϕˆ (z , t )⎥ (13)
ΔCCr ⎣ 0 ⎦
Now, the regulation error is defined by e = Y − Ysp = CCr − CCr − SP . For this case CCr −SP = constant ,
• ∂CCr ∂CCr ΔC
consequently e = . From mass balance equation, the uncertainty is ϕ = − u Cr .
∂t ∂t Δz
Now, substituting Equation (13) into it and taking the time derivative from the resulting equation, the
( )
• •• • t
(12) to obtain ϕˆ = τ 2 g1 + 1 e + (τ 1 − τ 2 g 2 ) e − g1
−1 −1 −1
τ 1e + τ 1 g 2−1 ∫ e(σ ) dσ . Integration of these equations
0
yields:
⎛ ⎞
ϕˆ = τ 2 (g1−1 + 1)e+ (τ 1 − τ 2 g 2−1 )e − ∫ g1−1τ 1e(σ ) dσ + ∫ ⎜⎜τ 1 g 2−1 ∫ e(σ ) dσ ⎟⎟ dζ
• t t
(14)
0 ⎝ 0 ⎠
The final expression for the input-output GMC controller with uncertainty estimation can be obtained
substituting the above equation into the non-ideal controller (Equation 13), to obtain:
Δz ⎡ −1 ⎛ −1 t ⎞ ⎤
( ) ( ) ( )
t • t
u= ⎢ 1g + (τ − τ g −1
) e + g −1
− g1 1 ∫
−1
τ e (σ )dσ − τ g −1
+ 1 e − ∫ ⎜
⎜ τ g ∫ e (σ ) dσ ⎟ dζ ⎥
⎟
ΔCCr ⎣⎢
1 2 2 2 2 1 1 2
0 0⎝ 0 ⎠ ⎦⎥
(15)
Note that this structure is equivalent to PID Controller plus double integral with time varying gains, where
the tuning of the controller’s gains depend of the closed-loop and measurement characteristic times of
the system.
192
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
Proposition 1. – Let ϕ and ϕ̂ be the uncertainty term and its estimate. The following dynamic system:
• • •
ϕˆ = τ 1 (ϕ − ϕˆ ) + τ 2 (ϕ − ϕˆ ) is an asymptotic-type reduced order observer for the system (9).
ε = ϕ − ϕˆ (16)
Now, the dynamic scalar equation of the estimation error according to Equation (9) and (12) is given by
• τ1 γ ( z, t )
ε =− ε+ (17)
1+τ2 1+τ2
⎛ τ1 ⎞ t ⎛ τ ⎞ γ ( z, t )
ε = ε 0 exp⎜⎜ − t ⎟⎟ + ∫ exp⎜⎜ 1 (t − s ) ⎟⎟ ds (18)
⎝ 1+τ2 ⎠ 0 ⎝1+τ2 ⎠ 1+τ2
⎛ τ1 ⎞⎛ Ψ⎞ Ψ
ε ≤ exp⎜⎜ − t ⎟⎟⎜⎜ ε 0 − ⎟⎟ + (19)
⎝ 1 + τ2 ⎠⎝ τ1 ⎠ τ 1
Ψ
ε ≤ (20)
τ1
Remark 1
It is important to analyze the structure of Equation (19) in order to obtain some characteristics of the
proposed observer. As it can be observed, taking the limit, when t → ∞ , the estimation error remains
Ψ
around closed ball with radius proportional to , this radius can be made as small as desired by taken
τ1
τ1 large enough. Besides, in order to improve the speed of the uncertainty observer convergence to the
steady-state estimation error, two actions have to be taken. The first one is to consider parameter τ1
large enough, which is necessary to obtain a small steady-state error, as mentioned above, and the
193
Journal of Applied Research and Technology
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
second one is faced with the influence of parameter τ 2 related to the derivative action of the uncertainty
observer. If τ 2 → −1 then exponential term of the right hand side of Equation (19) can be accelerated
enough and consequently, the convergence of the uncertainty observer exhibits a better performance.
In order to assess closed-loop stability, the analysis departs from the mass balance closed-loop equation
represented as follows:
∂e
= g1−1 e + g 2−1 e1 + ε
∂t
(21)
∂ e1
=e
∂t
Note that if not integral action were present (i.e. g2-1 = 0), the closed-loop system would be related to a
non homogeneous linear ordinary differential equation, with exogenous function ε (estimation error),
which is the standard observed based I/O linearizing control structure. One of the main drawbacks of this
methodology is the so-called peaking phenomenon (Sussman and Kokotovic, 1991), such phenomenon
can induce large overshoots which can lead to saturation of the control input; in addition, according to
the general description of the windup phenomenon described by Rönnbäck (1997), the control signal can
also be saturated due to large step changes in reference, even when a static control law is employed. In
such situation, feedback is broken and the plant behaves as an open-loop system with a constant control
input, allowing possible degradation of closed-loop performance. Now, considering the integral action
introduced, it acts as to compensate the estimation error term ε, which helps to avoid the mentioned
peaking and windup phenomenon.
∂E
= AE + Π (22)
∂t
Here:
Matrix A is Hurwitz stable by the adequate control’s gain selection. Solving (22) renders
t
E = E0 exp(− At ) + ∫ exp(− At )exp( Aσ )Π dσ (23)
0
194
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
A2. - If λ1, λ2,.., λk are the distinct eigenvalues of Matrix A, where λj has multiplicity nj and n1 + n2 + ... +
nk = n and ρ is any number larger than the real part of λ1, λ2,.., λk, that is ρ > max (ℜe(λj)), then there
exists a constant j > 0 that satisfies
exp( At )Ea ≤ j exp(− ρt ) Ea
Taking norms of Equation (23), the following inequality arises and considering A1 and A2:
⎡t Ψ ⎤
lim ⎢ ∫ j exp(ρ σ ) dσ ⎥
[
0 ≤ lim E ≤ E0 lim j exp − ρ ∫ dt +
t →t 0 t →t 0
(
t →t 0
⎢⎣ 0
)] τ1
lim j exp ∫ ρ dt ( )
⎥⎦
(24)
t →t0
∞
The above equation means that the case of uniform L´höpital´s rule can be applied as follows:
∞
Ψ
τ1
(
j exp ∫ ρ dt ) Ψ
0 ≤ lim E (t ) ≤ lim = lim
( )
(25)
t →t 0 t →t 0
j exp ∫ ρ dt ρ t →t 0 τ 1ρ
Ψ
E≤ →0 (26)
τ1ρ
As it is possible to note, it can be concluded that the regulation error lies on a closed ball of radius
proportional to the quota of the uncertain term.
Figure 4 shows results of the modeling of the tubular reactor compared to experimental data, note an
adequate performance of the proposed mathematical model. The considered parameters are D = 0.0115
m2/min; the initial length condition for the mass balance are CCr = 273mg / l at z = 0 ∀t and
∂ CCr
= 0 at z = 0 ∀ t , note that inlet concentration is considered as it is used in industry. The
∂t
initial time condition for the mass balance is taken as CCr = 50 mg / l for t = 0 ∀ z . The reactor’s
inlet velocity (momentum boundary condition) is vz = 0.18778 m / min for z = 0 ∀ t and the initial
condition is v = 0 m / min for t = 0 ∀ z , the bound of the control law was consider to be umin = 0
m/min and umax = 0.5 m/min.
195
Journal of Applied Research and Technology
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
150 150
a
b
120 120
90 90
60 60
30 30
0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
z z
150 150
c d
Cr(VI) Concentration [mg/L]
120 120
Cr(VI) Concentration [mg/L]
90 90
60 60
30 30
0 0
0 0.25 0.5 0.75 1 0 0.25 0.5 0.75 1
z z
Figure 4. Experimental data (■) and (__) numerical model results. (a) T = 4.68 min, (b) T = 5.14 min, (c)
T = 6.0 min, (d) T = 9.6 min.
Figures 5 and 6 show the open-loop and closed-loop dynamic behavior of the Cr(VI) concentration at
outlet reactor conditions. Note that the open-loop Cr(VI) trajectory reaches a steady state of 7.5 mg/l,
which is out of the maximum Cr(VI) concentration allowed (0.5 mg/l) by the environmental legislation; this
operating mode is obviously inadequate.
196
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
25
_._._ PID
20 ------- Linearizing
Cr(VI) Concentration [mg/l] ____ Proposed
15
10
0
0 10 20 30 40 50 60
Time [minutes]
0.7
Control Input [m/min]
0.6
0.5
0.4
0.3
0.2
0.1
0
0 10 20 30 40 50 60
Time [minutes]
In order to regulate the process, the observer based GMC (OBGMC) designed exhibits observer’s gains
fixed at τ1 = 2 min-1 and τ1 = 0.1 min-1; the resulting gains for the OBGMC are g1-1=0.1 min-1 and g2-1=0.1
min-1. For comparison purposes, a observer based I/O linearizing controller (OBLC) with an equal
proportional gain (g1- 1= 0.1 min-1) as OBGMC is considered and finally a linear PID controller is
implemented to regulate the Cr(VI) concentration; for the linear controller, an Input/Output identification
procedure was implemented with a step disturbance in the input flow of the order of the 5%, where the
corresponding steady state gain K = 47.92 (mg min) /(l m); the characteristic time τ = 7.7 minutes and
the time delay is θ = 1 minute. From these values, a minimum ITAE approximate model controller tuning
rules (Ogunnaike and Ray, 1994) was employed for the following controller’s gains gp = 0.1153 (l m)/(mg
min), τI =9.91 min-1, and τd = 0.356 min-1. Following a common industrial practice, the controller was
turned on at 35 minutes from the operation start.
Using the linear PID controller for the process, an oscillatory behavior is induced on the Cr(VI)
concentration being out of the legislation requirements, this behavior would be induced by the nonlinear
nature of the process and the associate lags for the distributed nature of the reactor. Control effort is also
unacceptable, thus, the performance of the linear PID control law looks very poor. With respect to the
OBLC, it can be seen the presence of large settling time; when the process starts, the estimation error is
not small enough and it leads to a far from satisfactory closed-loop performance of the corresponding
control effort. Finally, the proposed methodology acts keeping the controller at the minimum level for
reaching the large residence time allowed and, consequently, diminishing the Cr(VI) concentration at the
reactor’s outlet (Figure 6), more over Figure 7 is related with the time delay present on the control input,
which can induce some instabilities as mentioned above; however, for the proposed controller, it allows
to reach a satisfactory continuous operation to maintain an adequate volumetric flow.
0.55
___ Input
0.5 ----- Output
0.45
0.4
Velocities [m/min]
0.35
0.3
0.25
0.2
0.15
0.1
0.05
0 10 20 30 40 50 60
Time [minutes]
198
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
In order to show the impact of the regulation error, a performance index suggested by Ogunnaike and
Ray (1994) is the “Integral Time-Weighted Squared Error” (ITSE) defined by (27). ITSE exhibits the
advantage of heavy penalization of large errors at long time; therefore, it is a good measure of resilience
of the controller.
∞
ITSE = ∫ t ε 2 dt (27)
0
Figure 8 shows the corresponding performances for the different controllers employed, note the bad
behavior of the linear PID controller; on other hand, the OBLC shows a much better behavior, but the
proposed control law provides the best performance, it is related with the integral action included in the
control structure which, as it is well known, provides robustness.
10000
9000
_._._ PID
8000 ------- Linearizing
____ Proposed
7000
Performance Index [ITSE]
6000
5000
4000
3000
2000
1000
0
0 10 20 30 40 50 60
Time [minutes]
5. CONCLUSIONS
A tubular reactor for Cr(VI) reduction in plating wastewater is mathematically modeled for control
purposes by a dynamic plug flow with dispersion term. This model is experimentally validated. Employing
the distributed nature of the model (given by PDE), an observer based Generic Model Control is
proposed; this controller is equivalent to PID plus double integral actions with time varying gains. The
proposed methodology is robust against model uncertainties and provides a satisfactory closed-loop
performance in comparison to other control strategies.
199
Journal of Applied Research and Technology
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
6. REFERENCES
[1] Ricardo Aguilar-López and Rafael Maya-Yescas. State Estimation for Nonlinear Systems under
Model Uncertainties: A Class of Sliding-Mode Observers. J. Process Control 15, 363-370, 2005.
[2] R. Aguilar-López. Integral observers for uncertainty estimation in continuous chemical reactors:
differential-algebraic approach. Chemical Engineering Journal. 2003. 9, 113-120.
[4] Ricardo Aguilar-López and R. Maya-Yescas. State estimation for nonlinear systems under model
uncertainties: A class of sliding-mode observers. Journal of Process Control. 2005, 15, 363-370.
[5] APHA, AWWA, WPCF. (1995) Standard Methods for the Examination of Water and Wastewater,
19th Ed. American Public health Association, Washington, D.C., U.S.A.
Balas, M.J. Feedback control of linear diffusion processes. Int. J. of Control. 29, 523-533, 1979.
[6] Chistofides P.D. Robust control of parabolic PDE systems. Chem. Eng. Sci. 53, 2949-2965,
1998.
[7] Chistofides P.D., Dautidis, P. Robust control of hyperbolic PDE systems. Chem. Eng. Sci. 53, 85-
105, 1998.
[8] Department of Health and Human Services. Chromium toxicity. Agency for Toxic Substances and
Disease Registry, Division of Health Education and Promotion. Course: SS3048. 2000.
[9] Dochain, D, Babary, N, Tali-Mammar. Modeling and adaptive control of nonlinear distributed
parameter bioreactors via orthogonal collocation. Automatica. 28, 873-883, 1992.
[10] Froment, G.F., Bischoff, K.B. Chemical reactor analysis and design. Wiley, New York, 1990.
[12] Golub, D., and Oren, Y. Removal of Chromium from Aqueous Solutions by Treatment with
Porous Carbon Electrodes: electrochemical principles. [Link]. Electrochem, 19, 311-316, 1988.
[13] L. P. Lee and G.R. Sullivan. Generic Model Control. Computers and Chemical Engineering. 124,
573-580, 1988.
[14] Ogunnaike, B.A., Ray, W.H. Process Dynamics, Modeling and Control. Oxford University Press,
New York, 1994.
200
Vol. 6 No. 3 December 2008
Robust generic model feedback under model uncertainties: application of a tubular reactor for the treatment of industrial plating wastewater ,
R. Aguilar-López et al, 184-203
[17] M. G. Rodríguez, R. Aguilar, G. Soto and S. Martínez. Removal of chromium hexavalent from
rising chromatin waters by electrochemical reduction in a laboratory pilot plant. Water Science &
Technology. 49, 1, 115–122, 2004.
[18] Rönnbäck, S. Linear control of systems with actuators constraints. Ph D. dissertation. Lule
University of Technology, Sweden, 1997.
[19] G. Soto, R. Aguilar and S. Martínez. Numerical modeling of a no-steady state flow in tubular
reactor for treatment of industrial plating wastewater. Proceedings of the 5th IWA Chemical
Industry Group Conference. 306-314, November, 2002, Nimes, France.
[20] P.D. Signal and P.L. Lee. Generic model adaptive control. Chemical Engineering
Communications. 115, 35-52, 1992.
[21] Sussman, H.J., Kokotovic, P.V. The peaking phenomena and the global stabilization of nonlinear
systems. IEEE Trans. On Automatic Control. 36, 424-431, 1991.
201
Journal of Applied Research and Technology