Publication: CONTROL SUPERVISOR DE UNA CELDA JAMESON L-150 DE FLOTACIÓN PILOTO
Date
2016
Authors
ULLOA ESPINOZA, ANDRÉS IGNACIO
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
La presente investigación tiene por objetivo desarrollar e implementar un sistema de control supervisor en una celda Jameson de flotación piloto ubicada en el Laboratorio de Control de Procesos del Departamento de Ingeniería Civil Química de la Universidad Técnica Federico Santa María.Utilizando el sistema de control distribuido y simulador metalúrgico desarrollados en trabajos anteriores se implementa un sistema de control supervisor experto utilizando la razón de aire/pulpa y la profundidad de espuma como recursos para mantener la ley de concentrado de cobre entre el 27.7 y 28.7% maximizando la recuperación de cobre obtenida, teniendo un mínimo del 60%.El control supervisor cuenta con dos módulos, la rutina predictiva y la rutina experta retroalimentada, el primero utiliza un modelo simplificado para predecir los objetivos metalúrgicos obtenidos bajo ciertas condiciones de operación y alimentación, con lo que se pueden predecir los movimientos en las variables operacionales, utilizando la rutina experta retroalimentada para mantener los indicadores metalúrgicos dentro de los objetivos propuestos.El segundo es un control retroalimentado que corrige las desviaciones causadas por el primer módulo, éste elige el movimiento de uno de los dos recursos en base a un criterio de disponibilidad del recurso, para su acción el sistema debe estar en estado estacionario y el valor de proceso debe estar muy cercano a su set point.El modelo simplificado se construye utilizando el método de regresión de mínimos cuadrados parciales (PLS por sus siglas en inglés) mediante la herramienta PLS Toolbox de Matlab. El modelo utilizado en la rutina predictiva se construye en un rango de condiciones operacionales comunes logrando buenas predicciones generados usando un simulador metalúrgico.Finalmente, se estudian cuatro casos. En el primero se experimenta un aumento en la ley de alimentación de cobre y se evalúa el desempeño del control supervisor al habilitar y deshabilitar la rutina predictiva, los resultados son explícitos al mostrar que la acción de la rutina predictiva disminuye en 14 minutos el tiempo en que ambos objetivos son satisfechos respecto a un control sin la rutina habilitada.El segundo experimento muestra una disminución en la ley de alimentación de cobre, lo que genera la acción de la rutina predictiva. Debido a las condiciones extremas negativas de alimentación se obtiene una ley de concentrado dentro de la banda óptima y una recuperación muy por debajo del límite inferior. Para este caso se simula que el Scavenger, proceso hacia el cual se descargan las colas de la celda, no tiene capacidad de procesar adecuadamente una alimentación con las características actuales, por lo que mediante el panel de decisión se opta por activar solo el control de recuperación de la celda Jameson con un set point del 55% sacrificando la ley de concentrado de cobre.El tercer experimento muestra nuevamente una disminución en la ley de alimentación de cobre, sin embargo, para este caso se simula que el Scavenger puede aceptar una recuperación bajo el mínimo de la celda, por lo que el control supervisor se centra en mantener la ley de concentrado de cobre por sobre el mínimo para obtener el máximo de recuperación posible.El cuarto y último experimento coincide como la continuación de la experiencia anterior, donde se está operando en una condición poco favorable para el Scavenger y se alcanza un punto en donde es imposible mantener las condiciones actuales, por lo que se solicita al Rougher, proceso cuyo concentrado alimenta a la celda, un aumento en la ley de alimentación de cobre, en consecuencia, el control supervisor actúa en la dirección de maximizar la recuperación manteniendo la ley de concentrado de cobre dentro de la banda óptima.
The objectives of this research is to develop and implement a supervisory control system on a L-150 pilot Jameson flotation cell, available at the Process Control Laboratory of the Chemical Engineering Department, Santa María University, Valparaíso, Chile.Using a distributed control system and a metallurgical simulator, developed in previous projects, the supervisory expert control system is implemented to maximize the cell Cu recovery subject to produce a concentrate under technical specifications. These objectives are satisfied by modifying the set points of air to pulp ratio and froth depth of the distributed control system.The concentrate grade must be between 27.7 and 28.7% while the process Cu recovery has to reach at least 60%.The supervisory control system includes two main routines: expert feed forward control, to compensate for measured disturbances on the characteristics of the feed by using a simplified model to predict metallurgical objectives under the new scenery. The expert feedback control, takes care of the effect of other disturbances and some deviations produced by the first routine. This control act whenever a steady state is reached and selects the resource with lower saturation ratio.The simplified model is built applying the methodology if partial least square regression (PLS) implemented in the PLS Toolbox of Matlab. The data to fit the model was simulated by using the metallurgical simulator under normal operating conditions in order to achieve good predictions.Four cases are presented. In the first case, the response of the expert feedback control is compared with the response of both feedforward and feedback control when an increment on the feed Cu grade occurs. The results show that the action of both expert controls reduces in 14 minutes the time when both metallurgical objectives are satisfied.In the second case, a decrease of the feed grade is tested. After reaching a steady state the concentrate grade is upper the low limit while recovery is below the 60% limit. Since both objectives cannot be met simultaneously, the operator receives a message, to make a decision.In this case, it is assumed that the scavenger circuit is close to saturation and the operator decides to temporarily sacrifice the concentrate grade and increase the recovery target to 55%.In the third case, the same general conditions are kept and the feed Cu grade is decreased. Once again after the expert feed forward control acts, both metallurgical objectives cannot be met simultaneously and the operator receives a message. In this case, it is assumed that the scavenger has some room to process the actual Jameson cell tails. The expert feedback control continues adjusting the set points to maintain concentrate Cu grade above the low limit, allowing the recovery to decrease under 60%.The last case is the continuation of the previous one, but now the low recovery cannot be accepted because the scavenger circuit is close to reach its maximum capacity so the operator decides that the rougher concentrate Cu grade must be increased. After that change the supervisory control acts to get both objectives on target.
The objectives of this research is to develop and implement a supervisory control system on a L-150 pilot Jameson flotation cell, available at the Process Control Laboratory of the Chemical Engineering Department, Santa María University, Valparaíso, Chile.Using a distributed control system and a metallurgical simulator, developed in previous projects, the supervisory expert control system is implemented to maximize the cell Cu recovery subject to produce a concentrate under technical specifications. These objectives are satisfied by modifying the set points of air to pulp ratio and froth depth of the distributed control system.The concentrate grade must be between 27.7 and 28.7% while the process Cu recovery has to reach at least 60%.The supervisory control system includes two main routines: expert feed forward control, to compensate for measured disturbances on the characteristics of the feed by using a simplified model to predict metallurgical objectives under the new scenery. The expert feedback control, takes care of the effect of other disturbances and some deviations produced by the first routine. This control act whenever a steady state is reached and selects the resource with lower saturation ratio.The simplified model is built applying the methodology if partial least square regression (PLS) implemented in the PLS Toolbox of Matlab. The data to fit the model was simulated by using the metallurgical simulator under normal operating conditions in order to achieve good predictions.Four cases are presented. In the first case, the response of the expert feedback control is compared with the response of both feedforward and feedback control when an increment on the feed Cu grade occurs. The results show that the action of both expert controls reduces in 14 minutes the time when both metallurgical objectives are satisfied.In the second case, a decrease of the feed grade is tested. After reaching a steady state the concentrate grade is upper the low limit while recovery is below the 60% limit. Since both objectives cannot be met simultaneously, the operator receives a message, to make a decision.In this case, it is assumed that the scavenger circuit is close to saturation and the operator decides to temporarily sacrifice the concentrate grade and increase the recovery target to 55%.In the third case, the same general conditions are kept and the feed Cu grade is decreased. Once again after the expert feed forward control acts, both metallurgical objectives cannot be met simultaneously and the operator receives a message. In this case, it is assumed that the scavenger has some room to process the actual Jameson cell tails. The expert feedback control continues adjusting the set points to maintain concentrate Cu grade above the low limit, allowing the recovery to decrease under 60%.The last case is the continuation of the previous one, but now the low recovery cannot be accepted because the scavenger circuit is close to reach its maximum capacity so the operator decides that the rougher concentrate Cu grade must be increased. After that change the supervisory control acts to get both objectives on target.
Description
Catalogado desde la version PDF de la tesis.
Keywords
CONTROL DE PROCESOS , CONTROL EXPERTO , FLOTACION