An Neural Network Model for the Fe/SiO2 Ratio in Copper Flash Smelting Slag Using Improved Back Propagation Algorithm

Hong Lu

doi:10.4028/www.scientific.net/AMR.524-527.1963

Paper Titles

Numerical Simulation of the Swirl Flow Effect on the Cold Condition Characteristic of Entrained Flow Gasifier
p.1943

Research on the Extract Al₂O₃ from Coal Gangue
p.1947

Study on Recovering of Value Metals from Acid Leaching Solution of Roasted Electroplating Sludge
p.1951

The Influence of Electrolyte Heterogeneous Phenomenon on Tungsten Ion Exchange
p.1956

An Neural Network Model for the Fe/SiO₂ Ratio in Copper Flash Smelting Slag Using Improved Back Propagation Algorithm
p.1963

Energy Flow’s Consumption Analysis in Iron & Steel Factory
p.1967

Study on Methanol Synthesis with the Analog Components of Bluecoke Furnace Gas by Rich Oxygen Carbonization
p.1972

Microstructure and Properties of a Non-Quenched Prehardened Steel at Different Cooling Rate
p.1976

Analysis and Optimization of Material Flow and Energy Flow in the Coking-Steel Co-Production System
p.1980

HomeAdvanced Materials ResearchAdvanced Materials Research Vols. 524-527An Neural Network Model for the Fe/SiO2 Ratio in...

An Neural Network Model for the Fe/SiO₂ Ratio in Copper Flash Smelting Slag Using Improved Back Propagation Algorithm

Abstract:

The Fe/SiO2 ratio in slag is one of the important control parameters for copper flash smelting process, but it is difficult to describe the complex relationship between the technological parameters and the Fe/SiO₂ ratio in slag using accurate mathematic formulae, because the copper flash smelting process is a complicated nonlinear system. An neural network model for the Fe/SiO₂ ratio in copper flash smelting slag was developed, whose net structure is 8-15-12-1, and input nodes include the oxygen volume per ton concentrate, the oxygen grade, the flux rate, the quantity of Cu, S, Fe, SiO₂ and MgO in concentrate. In order to avoid local minimum terminations when the model is trained by back propagation (BP) algorithm, a new algorithm called GA-BP is presented by using genetic algorithm (GA) to determine the initial weights and threshold values. The results show that the model can avoid local minimum terminations and accelerate convergence, and has high prediction precision and good generalization performance. The model can be used to optimize the copper flash smelting process control.