基于粗选泡沫图像特征的浮选过程矿浆浓细度软测量.docVIP

  基于粗选泡沫图像特征的浮选过程矿浆浓 细度软测量 谢永芳，黄三思，桂卫华 （中南大学信息科学与工程学院，长沙 410083） 摘要：针对浮选过程矿浆浓细度难以实时检测的问题，提出一种基于 PLS 弱学习器的增强 回归集成软测量方法。该法基于获取的浮选泡沫图像特征数据，首先初始化 AdaBoost 指定 训练样本分布，通过调用弱学习器对训练集进行训练；然后根据训练结果更新训练集分布， 并再次训练，如此循环，从而建立浮选过程图像特征与矿浆浓细度之间的 Boost-PLS 关系模 型，实现矿浆浓细度的在线预测。实际生产数据仿真结果验证了所提方法的有效性、可行性。 关键词：浮选过程；软测量；浓细度；增强偏最小二乘回归算法 中图分类号：TP13 文献标志码：A  Soft sensor of the pulp density and fineness in flotation process based on flotation froth image features Xie Yongfang, Huang Sansi, Gui Weihua (School of Information Science and Engineering, Central South University, Changsha 410083, China) Abstract: Given the complexity of flotation process, it is difficult to accurately build the prediction model of pulp density and fineness. We propose a prediction method of Boosting-PLS which adopts the PLS as the basic learning algorithm in boosting for soft sensor of the pulp density and fineness. The algorithm is based on the bubble features which are extracted from the foam image by the digital image processing technique. The main idea of Boosting-PLS is to train a sequence of robust PLS models on various weighted versions of the original training set and then combine the predictions from the PLS models to obtain integrative results to establish a regression model about bubble features between the pulp density and fineness. Simulation results of actual production data in a flotation plant show the effectiveness and practicability of the proposed method compared to conventional multivariate statistical models. Key words: flotation process; soft-sensor; pulp density and fineness; Boosting-PLS 0 引言 浮选是利用浮选机、浮选柱等浮选设备，根据不同矿物表面亲水性和疏水性性质差异， 将有用矿物和脉石分离的过程[1]。浮选过程中，矿浆浓细度是衡量浮选性能的重要指标之一 [2-3] 。浓度过高，矿浆的充气度降低，而充气度的强弱又直接关系浮选时间，进而导致精矿 品位不高，过低时，矿浆流量随之增大，矿浆在浮选机中停留的时间就要缩短，回收率低； 细度过粗时，矿物颗粒不能单体解离，气泡难以被带起，同时在不断搅拌不易从气泡脱落， 使回收率提不高，过细时，虽然达到单体分离，但由于经过粉碎，形成矿泥多，又由于矿泥 会吸附大量药剂或毫无选择的吸附在粗粒矿物表面，而恶化了浮选过程，以致造成精矿品位 低，回收率下降，增加药剂消耗及磨矿费用等。由于目前的浓细度分析仪检测周期长，导致 矿浆浓细度无法在线测量，实际浮选生产过程，主要是采用人工方式进行离线检测，滞后性， 难以使浓细度值稳定在工艺指标的要求范围内。因此，研究