基于多源数据的铝土矿浮选生产指标集成建模方法

Integrated modeling for production index of bauxite flotation based on multi-source data

中文关键词  泡沫特征  生产指标  集成建模  偏最小二乘  正则极限学习机

英文关键词  froth features  production index  integrated model  partial least squares  regular extreme learning machine

中文摘要

      在长流程浮选过程中, 生产指标难以在线检测, 造成操作不及时, 影响系统的稳定运行. 本文提出了一种基 于多源数据的铝土矿浮选过程生产指标集成建模方法. 首先结合浮选机理和现场工人经验, 分析影响和反映生产 指标的多源数据(生产数据和泡沫图像特征数据); 然后分别建立各生产指标预测子模型和同步误差补偿子模型; 最 后采用信息熵和智能协调策略分别构建精矿品位和尾矿品位的集成预测模型. 工业验证和工况分析表明, 本文集 成建模方法具有良好的预测性能和较强的泛化性, 为基于生产指标的浮选过程操作参数控制和全流程优化奠定基 础.

英文摘要

      Because of the difficulty in online detecting the production index in a long-time flotation process, we are unable to take the counterpart operation timely to keep the production process stable. To deal with this problem, we propose an integrated prediction model for the production index of bauxite flotation based on multi-source data. Firstly, we make use of the flotation mechanism and the experiences of operators to analyze the multi-source data (production data and froth image features) that affects the production index. Then, the prediction sub-model and the corresponding compensation sub-model for each production index are built, and their parameters are optimized. Then, the integrated prediction models of concentrate grade and tailing grade are established by using the information entropy and coordination strategy. Industrial validation results show that the proposed method improves the prediction accuracy and adaptability, which lays a foundation for controlling the operating parameters of the flotation and the total process optimization.