全尾砂浆最佳絮凝沉降参数

摘要

为了得到最佳的絮凝沉降参数，研究使用BP 神经网络进行优化选择。通过对比分析，将输入因子简化为絮凝剂单耗和尾砂质量分数2个因子，输出因子简化为沉降速度1个因子；通过正交试验，建立网络学习、训练样本，优选出最佳网络模型。扩大正交试验，增加输入因子水平，组合优选样本，搜索最佳絮凝沉降参数。以司家营铁矿全尾砂絮凝沉降为例，优选出絮凝剂单耗为10 g/t，尾砂质量分数为18%，预测沉降速度为1.38 m/h，满足生产要求，比原生产所需絮凝剂单耗节省50%。应用结果表明：该研究成果效果显著，为絮凝沉降参数优选提供一种新思路。%Back-Propagation neural network was occupied in order to obtain the optimization of theflocculating sedimentation parameters. By practising comparison analysis,input data were simplified asthe flocculating agent consumptionand tailings concentration, and thesedimentation speed as the synthesized output data. By performing the numbered orthogonal tests, some learning and training samples were established so as to get the best network mode. Then, the best parameters were acquired using the selected network by expanding the orthogonal tests, increasing the levels of the parameters, optimizing the samplesand exploring the optimization of theflocculating sedimentation parameters. BP neural network mode was applied in Sijiayin Iron Mine.The results show that theflocculating agentconsumptionand tailingsmass fractionare10g/t, 18% respectively,andthesedimentation speed is1.38m/h, which meet the production requirementsand save 50% compared to the original production. The application indicates that this mode makes significant effect, providing a novel method toobtain the optimization ofthe flocculating sedimentation parameters.