An Expert System Designed to Improve Reliability of Assay Information in a Flotation Plant

"The objective of this expert system application in a copper-zinc concentrator, is to increase the availability of assay data through substitution for erroneous data. The assay data is used by the Operations personnel and automatic process control strategies to control flotation circuit performance. Two online X-ray fluorescence analysis systems analyze 24 streams across four divisions to provide the required data. Data availability of individual streams average 90% while the resulting circuit availabilities are 70%. The unavailability of correct data results from mechanical, X-ray or process problems. The erroneous data can be identified by process experts using knowledge relating stream interdependencies and historical process behaviour. An expert system was devised to provide real time identification of erroneous data and substitute for that data with the ultimate aim of improving metallurgical response by both the operator and automatic control systems.The paper discusses the early development work and architecture of the knowledge base designed using Neuron Data's expert shell NEXPERT.INTRODUCTIONThis paper describes the early development of an expert system (ES) prototype at the Kidd Creek concentrator. The project is a joint undertaking of the Kidd Creek Division of Falconbridge Limited and CANMET. Kidd Creek entered into the project to apply heuristic rules to the XRF assay data to improve the metallurgical control of the flotation process by increasing availability of analysis data. This project provided a problem of suitable size and future expandability while making use of a wide base of in-house expertise and experience. Another important benefit is the introduction of expert system technology to the site providing tools and experience for future system selection and technology application. CANMET has an on-going interest in assessing various aspects of ES technology (hardware, software, knowledge engineering, integration) while addressing problems in process operations that could most readily benefit from the technique. CANMET's role in the project was to provide project assistance, financial support, substitution algorithm assistance and software for data substitution. The knowledge bases (rules) have now been completed and are to be tested in the online environment."