Year 1997
Sampsa Laine

Ore type based expert system for the Hitura concentrator

At the Hitura mine changes in the mineralogy of the concentrator feed cause problems in process control. After a change in the feed type new process control method has to be found.
In this thesis a tree step methodology for finding proper classification variables using off-time information is presented. The first step is to traverse the concentrator history and determine the respective feed types. This is done by classifying laboratory analyses with the SOM. In this thesis the accuracy of the classification was studied by quantitative X-ray diffraction (XRD) and found satisfactory. The second step is to find on-line measurements that contain information about the type of feed. The third step is to formulate on-line SOM is calculated with these variables. The classification accuracy of the on-line SOM of this thesis was studied and found to be satisfactory.

Two methods for finding the appropriate treatment methods for feed types are presented. The first method is based on finding average treatments used for respective types of fees. The second one is based on collecting a large number of representative values of a feed type and forming a SOM with the treatments used. The treatments mapped into the elements of the SOM can be studied using arbitrary criteria.

The economical benefits of the expert system were estimated. The estimation was based on nickel lost in tailing and in MgO fines due to a high MgO content in the concentrate. The estimation suggested saving equal to three per cent of the Hitura turnover.

The applicability of the system to other plants and types of process was studied. The methodology of this system is portable to many types of process. However, development of the classification variables and the knowledge base may be laborious.

This info last modified 27 Apr 2024 - contact webmaster