锡石多金属硫化矿磨矿技术效率模型预测研究

研究了磨矿时间、磨矿浓度和磨机充填率对锡石多金属硫化矿磨矿技术效率的影响，利用Matlab编程技术建立了粒子群优化算法-BP神经网络磨矿技术效率预测模型并对模型进行了预测验证研究。结果表明，在磨矿时间8 min、磨矿浓度70%、充填率30%时，可获得较好的磨矿技术效率。迭代次数对锡石多金属硫化矿磨矿技术效率模型预测值与试验值误差影响显著; 在合适的迭代次数下，学习因子对模型预测值与试验值误差的影响很小，绝对误差小于±0.01个百分点，相对误差小于±0.04%。迭代次数达到500次后，粒子群优化算法-BP神经网络磨矿技术效率预测模型趋于稳定，模型可靠性高、适应性强。

Prediction Model for Grinding Technical Efficiency of Cassiterite-Polymetallic Sulfide Ore

The effects of grinding time, grinding concentration and medium filling rate on grinding technical efficiency of cassiterite-polymetallic sulfide ore were studied. Based on Matlab programming technique, a model of grinding technical efficiency prediction was established by using particle swarm optimization algorithm and BP neural network, and its prediction performance was verified. The results show that the best grinding technical efficiency can be obtained with the grinding time of 8 min, grinding concentration of 70% and medium filling rate of 30%. The number of iterations can bring significant effect to the error between the test value and the predicted value of the model of efficiency prediction. The learning factor has little effect on this error when the number of iterations is appropriate, showing the absolute error is less than ±0.01 percentage point and the relative error is less than ±0.04%. When the number of iterations is 500, the model of grinding technology efficiency prediction by particle swarm optimization algorithms and BP neural network tends to be stable, showing that the model is highly reliable and adaptable.