磁化处理的全尾砂料浆沉降参数优化模型

为了提高全尾砂料浆的脱水浓缩效果，将磁化处理技术引入到全尾砂料浆脱水浓缩中，并建立GA-SVM模型优选全尾砂料浆的沉降参数。建立支持向量机（SVM）沉降参数优化模型，以磁感应强度、磁化处理时间、料浆流速、料浆浓度、絮凝剂单耗为输入因子，沉降速度为综合输出因子，通过正交试验建立样本数据对SVM模型进行训练与检验，采用遗传算法（GA）对SVM模型参数进行优化，进而得到磁化全尾砂料浆沉降参数的GA-SVM优化模型。将GA-SVM模型运用到某铁矿磁化全尾砂料浆沉降参数优化中，得到的最佳沉降参数为磁感应强度0.192 T、磁化处理时间1.85 min、料浆速度1.92 m/s、PAC单耗28 g/t，沉降速度可达约155 cm/h。研究表明：适宜的磁化处理条件可提高全尾砂料浆的脱水浓缩效果，节约30.0%~42.5% PAC用量，GA-SVM模型对全尾砂料浆沉降参数预测结果相对误差在5%以内、预测精度高，为全尾砂料浆脱水浓缩及其参数优选提供了一种新思路。

Optimal prediction model on sedimentation parameters of pre-magnetized crude tailings slurry

In order to improve the dewatering and concentrating effect of crude tailings slurry (CTR), magnetic treatment technique was introduced into dewatering and concentrating of CTR, and a GA-SVM(genetic algorithm-support velocity machine) model was established to optimize sedimentation parameters of CTR. The SVM model for predicting sedimentation parameters of pre-magnetized CTR was established and trained with sample data got from orthogonal experiments, taking magnetic induction, magnetized time, cycling velocity and mass concentration of CTR, unit consumption of flocculant as input factors, and sedimentation velocity as comprehensive output factor. Then a GA-SVM model for optimizing sedimentation parameters of pre-magnetized CTR could be obtained after parameters of SVM model optimized by the genetic algorithm. Furthermore, the GA-SVM model was adopted into an iron mine''s pre-magnetized CTR to optimize its sedimentation parameters, and the optimized sedimentation velocity was about 155 cm/h when magnetic induction, magnetized time, cycling velocity of CTR and unit consumption of PAC flocculant was 0.192 T, 1.85 min, 1.92 m/s and 28 g/t, respectively. The study results show that dewatering and concentrating effect of CTR could be improved and PAC could be saved by 40% under suitable condition of magnetic treatment. For optimizing sedimentation parameters of pre-magnetized CTR, the relative prediction error of GA-SVM model was less than 5%, which suggests that the GA-SVM model has higher prediction precision. The study also providing a new method to CTR''s dewatering and concentration as well as its parameter optimization.