小波分析是一种新的信号处理技术，具有良好的时频局部化特征。为了克服BP网络自身算法的缺陷，得到更高的学习精度和更快的收敛速度，使用小波包分析的特征提取及神经网络的非线性映射特性，构造了小波神经网络，以此为基础开发出的软件系统具有使用的特征量少，建造预报系统较为简单等优点。将之应用于炼铜转炉炉渣重量及成分预报，该模型完全能够较准确地预报出渣量和成分。其平均拟合误差为1.5％，平均预报误差为3.1％。

Based on the material balance and heat balance，300 groups of historical data about copper converters are simulated dynamically on the computer and main element composition of a melt and the temperature in the molten bath are calculated every other one minute in the period of air refining.Then，they are fitted with quartic curve and the coefficients gotten by fitting are regarded as outputs of neural network.Partial initial data and the optimized operation parameters are considered as inputs of the neural network.Through training，the neural network can simulate the change of the furnace condition during air refining.It is able to self-study by increasing the number of new samples.It is applied to monitor the furnace condition of copper converter and constitute a real-time forecast system of furnace condition in copper converter.The system has been successfully used in operation of copper converter in a smelter.Production quota is obviously improved after running-in for four months.Copper production and the weight of cold input are increased respectively by 6.0％ and by 8.0％.Average converter life-span is improved from 213 to 235 times.A new idea has been provided for realization of in-line real-time control of copper converter.

To control the smelting temperature of the copper converter((1 250±10) ℃ in slag-making period, (1 180±10) ℃ in copper-making) and improve the life-cycle of the converter, the smelting process of cold materials is simplified in this paper and the dynamic models of smelting rate and smelting time are established. The parameters of the models are determined. The models are used in the Intellectual Decision Support System of the Operation- Optimization of copper converter. The software system automatically tells operators the maximum feeding of each cold material according to the current left heat and the melt temperature showed in the monitor interface of furnace conditions. Thus, the melt temperature can be controlled within a feasible limits. The practice shows that the application of the models makes the average life-cycle of the converter lining be improved from 213 furnace-times to 235 furnace-times and make the cold materials feeding improved by 8 percent under the condition that not enough cold materials are provided.