Operational Pattern Matching-Based Key Technical Indexes Prediction For Copper Flash Smelting Process

Abstract:Flash smelting is the main metallurgy technique to smelt copper. The key technical indexes that indicate the comprehensive condition in the smelting process are usually measured by operators with time delays, which can not reflect the condition level in real time. Moreover, a large number of production operation data which reflects the condition information is accumulated in the smelting process. Similar data with current condition can reflect the operation law of current condition. Therefore, the prediction of three key technical indexes in the process of copper flash smelting is studied, which are matte temperature, matte grade and mass rate of Fe and SiO2, so as to realize the monitoring of melting condition in real time.In this paper, technical process of copper flash smelting and main factors which affect the smelting condition are analyzed. On this basis, a general framework of the key technical indexes prediction model based on operation pattern matching is constructed. The main research contents and innovative points in this paper are as follows.(1) Considering the problem of low speed and low accuracy of similar operational pattern retrieval caused by the enormous operational pattern library, two level matching method of operational pattern is presented for copper flash smelting process. Comprehensively considering spatial structure of the sample data, the optimal attributes are chosen by probability density approximation principle, and attribute importance are measured with information entropy, which quickly eliminates the patterns that cannot meet the requirements, so the roughing operational patterns are obtained. Then geodesic distance is used as similarity measurement method in the secondary pattern matching, so the similar operational patterns are obtained. Under the premise of guaranteed matching accuracy, it effectively improves the speed of matching.(2) Considering the measured lag of the key technical indexes and difficulties in choosing training samples caused by enormous data in the smelting process, the prediction method of the key technical indexes is presented for copper flash smelting process. The similar operational patterns is extracted by pattern matching technique, with the projection pursuit regression method, and a prediction model of the key technical indexes based on projection pursuit regression is constructed. The projection direction of the prediction model is optimized and updated using accelerating genetic algorithm, which effectively improves the prediction accuracy of key technical indexes.(3) Combined with the actual production process for copper flash smelting of a smelter, using the research results in this paper, a monitoring system based on the operational pattern is developed for copper flash smelting process. The results of actual operation are given to verify the effectiveness and feasibility of the proposed method. Figure25, table5, reference71.