Development of a Machine Vision System to Monitor a Grinding Mill Prototype Timo Roinea, Janne Pietiläa, Jani Kaartinena, Peter Blanzb, Jörn Rohlederc, Pertti Rantalac a
Helsinki University of Technology (TKK), Department of Automation and Systems Technology, P.O.Box 5500, FI-02015 TKK, Finland (e-mail:
[email protected],
[email protected],
[email protected]) b
Outotec Oyj, Minerals Processing division, P.O.Box 84, FI-02201 Espoo, Finland (e-mail:
[email protected]) c
Helsinki University of Technology (TKK), Department of Materials Science and Engineering, P.O.Box 6200, FI-02015 TKK, Finland (e-mail:
[email protected],
[email protected])
Abstract A grinding mill model based on discrete element method (DEM) simulations is being developed at Outotec Oyj (Finland) to be used in mill design optimization. The model can be used for many purposes; one example is the selection of the lining and the size of the mill to meet the requirements of the clients. To validate the accuracy of the DEM simulator, a laboratory-sized ball mill prototype was constructed and iron balls were used as the mill charge. The prototype mill has its front side made of transparent glass in order to be able to visually examine the behaviour of the ball batch while the mill is rotating. The idea behind this type of arrangement is to create an exact model of the prototype mill for the DEM simulator and then compare the results while running the simulator and the physical mill with identical filling and rotation speed parameters. Image analysis can be used to evaluate and compare the performance of the prototype mill and the DEM simulator. Because of the need to process a lot of images, a machine vision system was constructed. The system was used to analyze the desired properties from the images taken of the physical prototype mill and from the images acquired from the virtual DEM simulator model of the same mill. By using the developed image analysis software, the ball batch can be successfully separated from the background, despite the challenges caused by dust build-up, which influences the visibility of the balls. Several measures, such as shoulder and toe angle and the form of the upper edge of the ball batch, are calculated from both camera and DEM based images. Because the measurements are calculated identically in both cases, they can be compared to study and to improve the accuracy of the simulator.
Keywords grinding, mill, mineral processing, machine vision, image analysis, discrete element method
1. Introduction In mineral processing industry, the valuable minerals in the ore need to be liberated from the unwanted gangue by comminution before they are separated (by flotation, for example). Also, the particle size needs to be manipulated to be suitable for the separation method used. The comminution begins with crushing of the ore and continues usually with grinding, resulting in a particle size with relatively clean particles of mineral and gangue. The grinding is typically performed in a rotating cylindrical mill, which contains a charge of crushing bodies moving freely inside the mill and thus comminuting the ore particles. Comminution is usually performed with water, but in certain applications also dry grinding is used. Grinding can be performed by different mechanisms, which include impact (or compression), chipping, and abrasion as presented in Figure 1. Impact grinding happens due to forces applied almost normally to the particle surface, chipping is caused when the forces are oblique and abrasion when they are parallel. (a)
(b)
(c)
Figure 1. Different mechanisms of comminution: (a) impact or compression, (b) chipping, (c) abrasion. The grinding mills can be classified into tumbling mills and stirred mills according to the method by which the motion is induced to the charge. Tumbling mills are typically used in coarse grinding, resulting in a typical particle size between 40 and 300 μm while stirring mills are used in fine (15-40 μm) and ultra-fine (