面向泡沫特征形态分布的矿物浮选过程加药量控制

The reagent dosage is the key operation variable for mineral flotation, which has a critical influence on flotation indices and effective recovery of mineral resources. Combined with China's actual situation of mineral ore with low grade and complex composition, the operating mode of reagent dosage that the workers inspect and observe bubble status and the reagent dosage manually is frequently regulated is difficult to meet stable running of production in the flotation plant. So，based on the viewpoint that the form distribution of bubble visual characteristic is direct indicator of flotation performance and closely related to the reagent dosage, the select of the key bubble characteristic and the estimation of its' form distribution will be investigated according to flotation production performance. The relational model between form distribution of bubble visual characteristic and reagent dosage need to be investigated and form distribution of bubble characteristic based reagent dosage control method is proposed. On the basis, a pulp nature based feedforward compensator will be investigated because the reagent dosage is not promptly regulated when the flotation nature change largely. In view of strong relevance of the cascade process, the sampling sites of sub-process will be selected for flotation process, and association process based cascade control method for reagent dosage is explored in order to be capable to quickly and accurately adjust reagent dosage for cascade processes. It has an important practical significance that the successful implementation of this project will improve flotation indices and lower resource consumption under the condition of existing equipment and process for domestic flotation plants.

加药量是矿物浮选生产过程关键操作量，其控制合理与否直接影响浮选工艺指标的稳定与矿物资源的有效回收。在成分复杂、低品位矿源的国情下，国内选厂实行“人工巡视观泡、手动频繁调药”的加药量操作模式难以保证浮选生产的稳定优化运行。为此，本项目基于“泡沫视觉特征形态分布是浮选性能的直接指示器且与加药量密切相关”的视角，研究基于工况变化的关键泡沫特征选择及其形态分布有效表征方法；探讨泡沫特征形态分布与加药量的耦合关系模型方法，提出面向泡沫视觉特征形态分布的加药量控制方法；在此基础上，针对入矿条件变化导致加药量调整不及时的问题，构建基于入矿条件变化的加药量前馈补偿器；针对浮选工序关联性强，拟在各工序设置采样点，研究基于工序关联的浮选全流程加药量级联控制方法，以快速准确地调整级联工序加药量，优化浮选生产。本项目成果对于国内浮选企业在现有设备和工艺流程条件下提高生产指标、降低资源消耗具有重要现实意义。

在成分复杂、低品位矿源的国情下，国内选厂实行“人工巡视观泡、手动频繁调药”的加药量操作模式难以保证浮选生产的稳定优化运行。为此，项目组首先提出了基于稀疏化神经网络的特征选择方法和基于Hessian正则的自适应损失半监督稀疏特征选择算法以提取泡沫关键特征；针对环境恶劣、光衰等现象容易导致颜色特征失色，讨论了基于Grey Edge参数主动优化的光照补偿算法；针对深度学习具有学习特征的特性，提出了基于CNN-SVM的矿物浮选故障诊断方法；然后提出基于静态设定和动态补偿的药剂量优化控制方法，以实现复杂工业过程生产指标的动态优化控制；接着讨论基于B样条的泡沫特征PDF表征方法，采用MLS-SVM建立了基于泡沫特征PDF和生产指标的模型；最后提出了基于泡沫特征PDF的药剂量预测控制；针对浮选流程长的特点，提出了基于模糊的前馈补偿策略。本项目成果对于国内浮选企业在现有设备和工艺流程条件下提高生产指标、降低资源消耗具有重要现实意义。