Intelligent Decision Support System for Alumina Point Feeder Inspection and Tending

First we identified and tested instruments for imaging, 3D measurements, spectroscopy, temperature and thermography. The aim of the tests was to find which measurement techniques are most suitable for further development towards the project goal The tests showed which instruments are best suited for detection and quantification of various types of problems in the feeding area. A platform (hardware and software) was assembled, forming the basis for further development. The main aims of the project were divided into various subparts, each of which was addressed individually. At first the main focus was on detecting cladding on the hammer which breaks the crust. Camera measurement techniques combined with lasers was used for this purpose. The same technology was later used to detect the oxide level and crust openings.

Aluminium is produced by electrolysis of aluminium oxide (alumina). Alumina as powder is added to the electrolyte automatically. This is done by a socalled point feeder technology, using a pneumatic hammer breaking the frozen top crust. The alumina is fed to the feeding area in a closed system. Process disturbances may upset the feeding operation leading to reduced productivity from loss in current efficiency and increased emissions of fluorides, dust, PAH and climate gases. It is imperative to have a clo se follow-up of the feeding area in order to discover any deviations at the earliest possible time to take corrective actions. Both the follow-up in the form of daily manual inspections and corrective actions are done by operators. The types of corrective actions are defined based on how the condition differs from the standard. At the Lista plant, there are 276 cells having two point feeders each. To inspect all point feeders, operators have to move a total distance of 6 km. Hence routine inspections a re costly and time consuming. The point feeders should be inspected several times per day. Feeders with operating problems require additional inspections and corrective actions. The aim of the project is to develop a tool which can perform this inspecti on automatically, and determine the condition of the point feeder area and corrective actions. This tool will in time be mounted on a robot moving from cell to cell. The robot will obtain access to the closed feeding area. An automatic system for point feeder inspection will enable a much more frequent, accurate and objective inspection, compared to the current situation. This will lead to a more stable operation with higher productivity, less emissions and less maintenance.