贵州某磷尾矿充填体有害成分溶出浓度预测研究

Study on prediction of extraction concentrations of harmful components in phosphate tailings backfill in Guizhou

本文借助补偿模糊神经网络,以贵州某磷尾矿充填体溶出试验数据作为学习样本,建立溶出条件与溶出质量浓度间的数学模型,通过误差―步数曲线及网络验证可知:针对溶出试验中F-和P溶出质量浓度建立的2个4输入单输出网络模型收敛性均好,最大预测误差分别为-4.86%和5.71%,平均预测误差分别为-0.945%和0.985%,能起到较好的预测作用。

By means of a compensation fuzzy neural network, the test data of dissolution phosphate tailings backfill of a phosphate mine in Guizhou are used as learning samples and the mathematical model between dissolution conditions and the concentration of dissolution are established. The results of the step-error curves and network verification show that the two four-input and singleoutput network structure models, which focus on the dissolved concentration of F-, P , have good behavior of convergence. The maximum prediction error are -4.86% and 5.71% respectively and the mean prediction error are -0. 945% and 0. 985% respectively, which can have a significant predictive effect.