辉钼矿回收率对水体磁化改性因素敏感性的BP神经网络研究

为优化磁化水体系中辉钼矿回收指标,以正交实验的方法进行了辉钼矿单矿物浮选试验,并利用BP神经网络对辉钼矿回收率与水体磁化改性影响因素的敏感性关系进行分析。结果表明:浮选过程中辉钼矿回收率对不同磁化因素的敏感性由大到小依次为:电流频率、退磁时间、磁化时间、电流频率与磁化时间交互影响和电流频率与退磁时间交互影响。研究利用电流频率等三个主要影响因素,通过BP神经网络作为模型对辉钼矿的回收率进行预测,其拟合度与精度较好,拟合优度R^(2)为0.9704,相对平均误差仅为1.27%,该模型能较好地用于辉钼矿回收率的预测。研究对利用磁化水改善金属硫化矿浮选效果的工业应用有一定的参考意义。

The sensitivity of molybdenite recovery to factors of water magnetization modification by BP neural network

In order to optimize the recovery index of molybdenite in the magnetized water system, a single-mineral flotation test of molybdenite was conducted by orthogonal experiment, and the sensitivity relationship between the recovery of molybdenite and the influencing factors of magnetization modification in water was analyzed by BP neural network. The results showed that the sensitivity of molybdenite recovery to different magnetization factors during flotation was in descending order: current frequency, demagnetization time, magnetization time, interaction between current frequency and magnetization time, and interaction between current frequency and demagnetization time. The study used the three main influencing factors such as magnetic induction strength to predict the recovery of molybdenum by the BP neural network action model, which has a good fit and accuracy, with a good fit R2 of 0.9704 and a relative average error of only 1.27%, and the model can be better used for the prediction of the recovery of molybdenite. The study has some reference significance for the industrial application of using magnetized water to improve the flotation effect of metal sulfide ores.