Thesis Análisis, calibración y validación de modelos de potecia para molienda SAG en molinos de 32’ X 15’
Loading...
Date
2019-04
Authors
Journal Title
Journal ISSN
Volume Title
Program
Ingeniería Civil Metalúrgica
Campus
Campus Casa Central Valparaíso
Abstract
El procesamiento de minerales de cobre sulfurado debe ser realizado en instalaciones de chancado y molienda para posteriormente ser flotado y producir el concentrado de cobre. Para el proceso de molienda se utilizan molinos horizontales, los que se clasifican en autógenos (AG), semi-autógenos (SAG) y de bolas.
Para maximizar el rendimiento del proceso y reducir el impacto de los costos, es necesario aumentar la producción del molino SAG y operar en el punto óptimo económico determinado por un nivel de carga y bolas específico. Para predecir el nivel de bolas se utilizarán los modelos de potencia, los que permiten generar un control predictivo de la operación y realizar un seguimiento diario en base a datos representativos del día, con el fin de conocer las condiciones operacionales y generar correcciones pertinentes para obtener el óptimo operacional.
El objetivo de esta memoria es analizar, calibrar y validar modelos de potencia en un molino SAG de 32’ x 15’. Para efectos de cálculo operacional, se consideraron los modelos de Arbiter-Harris, Austin, Bond, Hogg-Fuerstenau y Morrell, ya que utilizan variables medidas a diario en la operación. Para la calibración se busca minimizar el error entre la potencia real y la modelada, los datos utilizados fueron obtenidos mediante Crash Stop y Grind Out. Para efectos de implementar los modelos a diario y mantener el control, se generaron relaciones metalúrgicas que permiten establecer el nivel de carga al interior del molino. Se validan los modelos utilizando el error absoluto con respecto al nivel de bolas real, utilizando dos métodos. El primero consiste en utilizar la constante de ajuste anterior a la calibración, y el segundo, al promedio acumulado de constantes previas a calibración. Los datos utilizados son propiedad de Codelco Chile y CodelcoTech.
Se concluye que los modelos estudiados son aplicables operacionalmente, la clasificación de modelos con menor error absoluto respecto al nivel de bolas real está dada por Hogg-Fuerstenau (3,4%), Austin (4,3%), Bond (5,0%), Arbiter-Harris (5,4%) y Morrell (8,1%), utilizando el método que promedia las constantes de calibración. Se recomienda implementar el modelo de Hogg-Fuerstenau para realizar el seguimiento diario dada su simpleza y fácil aplicación.
The processing of sulphide copper minerals must be carried out in crushing and milling facilities to be floated and produce the copper concentrate. Horizontal mills are used in the milling process, which are classified as autogenous (AG), semi-autogenous (SAG) and ball mills. To reduce the cost impacts, the performance of the process has to be maximized, it is necessary to increase the production of the SAG mill and operate at the optimum economic point which is determined by a specific charge and grinding media level. Power draw models will be used to predict the grinding media level, which allows a predictive control of the operation based in a representative set of daily data, with the purpose of verifying the operational conditions and to fix any condition to achieve the optimum operational point. The objective of this report is to analyze, calibrate and validate power models in a SAG mill of 32’ x 15’. For operational calculation purposes, Arbiter-Harris, Austin, Bond, Hogg-Fuerstenau and Morrell models were considered, since they use variables measured daily in the operation. For calibration, the method used was error minimizing between the real and modeled power, data obtained by means of Crash Stop and Grind Out were used. For the purpose of implementing the models daily and maintaining control, metallurgical relationships were generated, which allows to establish the level of charge inside the mill and then predict the level of balls. The models were validated using the absolute error with respect to the actual ball level, using two methods. The first one consists in seeing the prediction capacity of the adjustment constant prior to calibration, and the second, the accumulated average of constants prior to calibration. The data used is property of Codelco Chile and CodelcoTech. It is concluded that the models studied are operationally applicable, the classification of models with the lowest absolute error respect to the actual ball level is given by Hogg-Fuerstenau (3,4%), Austin (4,3%), Bond (5,0 %), Arbiter-Harris (6,4%) and Morrell (8,1%), using the method that averages the calibration constants. It is recommended to implement the Hogg-Fuerstenau model to carry out daily monitoring given its simplicity and easy application.
The processing of sulphide copper minerals must be carried out in crushing and milling facilities to be floated and produce the copper concentrate. Horizontal mills are used in the milling process, which are classified as autogenous (AG), semi-autogenous (SAG) and ball mills. To reduce the cost impacts, the performance of the process has to be maximized, it is necessary to increase the production of the SAG mill and operate at the optimum economic point which is determined by a specific charge and grinding media level. Power draw models will be used to predict the grinding media level, which allows a predictive control of the operation based in a representative set of daily data, with the purpose of verifying the operational conditions and to fix any condition to achieve the optimum operational point. The objective of this report is to analyze, calibrate and validate power models in a SAG mill of 32’ x 15’. For operational calculation purposes, Arbiter-Harris, Austin, Bond, Hogg-Fuerstenau and Morrell models were considered, since they use variables measured daily in the operation. For calibration, the method used was error minimizing between the real and modeled power, data obtained by means of Crash Stop and Grind Out were used. For the purpose of implementing the models daily and maintaining control, metallurgical relationships were generated, which allows to establish the level of charge inside the mill and then predict the level of balls. The models were validated using the absolute error with respect to the actual ball level, using two methods. The first one consists in seeing the prediction capacity of the adjustment constant prior to calibration, and the second, the accumulated average of constants prior to calibration. The data used is property of Codelco Chile and CodelcoTech. It is concluded that the models studied are operationally applicable, the classification of models with the lowest absolute error respect to the actual ball level is given by Hogg-Fuerstenau (3,4%), Austin (4,3%), Bond (5,0 %), Arbiter-Harris (6,4%) and Morrell (8,1%), using the method that averages the calibration constants. It is recommended to implement the Hogg-Fuerstenau model to carry out daily monitoring given its simplicity and easy application.
Description
Keywords
Industria minera, Molinos y molienda, Diseño y mantención de maquinarias