TY - THES

T1 - Predicting Cone Crusher Performance Using Regression Analysis

AU - Eliasson, Johan Eric

N1 - embargoed until null

PY - 2013

Y1 - 2013

N2 - The objective of this research has been to develop a new method and a model that can be used to predict the capacity and the product particle size distribution that are typically produced by a hydrocone cone crusher. Today the method for calculating capacity and product particle shape distribution is time consuming and often requires that complete 3D model of the chamber before it can be calculated. The main value of this new method is to quickly gain an insight to what capacity and product size distribution a cone crusher will produce, in other words what size of crusher is needed to fulfill set requirements. The model is simple, it is built on many different tests taken in the laboratory in Svedala and it is a first approach to add more crusher specific parameters to the calculations of capacity and product size distribution. The test center at Sandvik SRP in Svedala receives raw material samples, product samples and feed samples from all over the world. The data from all tests conducted are saved in a database. The first part of the project has been to construct an interface between this database and MATLAB where all further calculations have been made. After the interface is set up all the different parameters needed to build the model can be fetched automatically. The interface sorts all samples and only the samples passing all set criteria are later used for calculations. The model has 10 input parameters taking feed, chamber geometry, crusher specific settings etc. into consideration and the 4 output parameters of the model is a predicted capacity as well as 3 parameters needed to construct a product particle size distribution using the Swebrec formula. The model should be considered to be in its alpha stage; R2 values for the particle size distribution varies from 0.65 – 0.94 and the R2 value for the capacity prediction is 0.91. Future work on improving the model should include more samples taken from the field with the model specifically in mind, more field tests with different kinds of material and more samples taken with different crusher settings. The correlation between the different input parameters and their effect on the final product should also be more thoroughly investigated.

AB - The objective of this research has been to develop a new method and a model that can be used to predict the capacity and the product particle size distribution that are typically produced by a hydrocone cone crusher. Today the method for calculating capacity and product particle shape distribution is time consuming and often requires that complete 3D model of the chamber before it can be calculated. The main value of this new method is to quickly gain an insight to what capacity and product size distribution a cone crusher will produce, in other words what size of crusher is needed to fulfill set requirements. The model is simple, it is built on many different tests taken in the laboratory in Svedala and it is a first approach to add more crusher specific parameters to the calculations of capacity and product size distribution. The test center at Sandvik SRP in Svedala receives raw material samples, product samples and feed samples from all over the world. The data from all tests conducted are saved in a database. The first part of the project has been to construct an interface between this database and MATLAB where all further calculations have been made. After the interface is set up all the different parameters needed to build the model can be fetched automatically. The interface sorts all samples and only the samples passing all set criteria are later used for calculations. The model has 10 input parameters taking feed, chamber geometry, crusher specific settings etc. into consideration and the 4 output parameters of the model is a predicted capacity as well as 3 parameters needed to construct a product particle size distribution using the Swebrec formula. The model should be considered to be in its alpha stage; R2 values for the particle size distribution varies from 0.65 – 0.94 and the R2 value for the capacity prediction is 0.91. Future work on improving the model should include more samples taken from the field with the model specifically in mind, more field tests with different kinds of material and more samples taken with different crusher settings. The correlation between the different input parameters and their effect on the final product should also be more thoroughly investigated.

M3 - Master's Thesis (University Course)

ER -