基于拉曼光谱的矿物浮选精矿品位在线检测方法

Concentrate grade is an important technology indicator of mineral flotation process, which can directly reflect operation state of flotation process and product quality. Therefore, it has become an important indicator for operation optimization and feedback control of flotation process. On-line detection of concentrate grade is a difficult problem to be solved in mineral flotation process, which has become a bottleneck restricting the improvement of the level of mineral flotation automation. Focusing on the difficult problem of the online detection of Raman spectra of mineral in foam layer through concentration process, a novel online detection method by Raman spectroscopy will be studied. Considering the difficulties, including the on-line measurement of Raman spectrum from the foam layer in the concentrate process under complex environment, the Raman spectrum superimposed with strong fluorescence scattering signals produced by impurity minerals, overlapped peaks between the characteristic peaks of Raman spectra of different minerals and obtaining precise quantitative models for concentrate grade under cross interference based on Raman spectrum, an on-line measuring method of Raman spectrum from the foam layer will be investigated, a strong background subtraction method for Raman spectrum will be studied. And then, a characteristic wavelength selection method based on peak features of Raman spectrum will be investigated. Finally, a nonlinear modeling method based on Raman spectrum under cross interference will be studied. The research will break through key technologies for on-line detection of concentrate grade in mineral flotation process by Raman spectroscopy. The achievements will provide a new way to on-line detection of important production indicators in mineral flotation process. More importantly, it will promote the automation of mineral flotation process in China. Therefore, it is of great significance for both scientific research and industrial application.

精矿品位是矿物浮选过程的关键生产指标，能够直接反应浮选过程的运行状态和产品质量，成为浮选过程运行优化和反馈控制的重要参数。精矿品位的在线检测是矿物浮选过程中亟待解决的难题，成为制约矿物浮选自动化水平提升的瓶颈。本项目围绕矿物浮选中精矿品位在线检测难题，研究提出一种基于拉曼光谱的新的在线检测方法。针对精选工序中泡沫层矿物的拉曼光谱在线采集困难、杂质矿物强荧光信号叠加、不同矿物之间拉曼光谱特征峰重叠以及交叉干扰下精矿品位定量模型构建等问题，研究精选工序泡沫层矿物的拉曼光谱在线获取方法、拉曼光谱强荧光背景扣除方法、融合谱峰信息的拉曼光谱特征波长筛选方法、交叉干扰下拉曼光谱非线性建模方法，突破基于拉曼光谱的矿物浮选精矿品位在线检测关键技术。研究成果将为矿物浮选关键生产指标的在线检测提供新思路，并将促进矿物浮选过程自动化水平的提升,具有重要的理论意义和应用价值。

项目背景：精矿品位是钨浮选过程的关键生产指标，能够直接反应浮选过程的运行状态和产品质量，成为浮选过程运行优化和反馈控制的重要参数。实现精矿品位在线定量分析是实现钨浮选自动化水平提升的关键步骤。本项目围绕钨浮选中精矿品位在线检测难题，提出一种基于拉曼光谱的矿物成分在线定量分析方法。针对复杂环境下钨精选泡沫层矿物的拉曼光谱提取困难、杂质矿物强荧光信号叠加以及交叉干扰下精矿品位定量模型构建等问题，研究泡沫层拉曼光谱在线高效提取方法、拉曼光谱强荧光背景扣除方法和交叉干扰下拉曼光谱非线性建模方法，突破基于拉曼光谱的钨浮选精矿品位在线检测关键技术。研究成果将为钨浮选关键生产指标的在线检测提供新思路，并将促进钨浮选过程自动化水平的提升, 具有重要的理论意义和应用价值。.主要研究内容：为了验证方法的有效性，在三年的研究期间，主要开展了矿物浮选过程中原矿多组分快速识别方法，精选工序中泡沫层矿物的拉曼光谱在线获取方法，拉曼光谱强荧光背景动态扣除方法，浮选过程矿物拉曼光谱特征谱峰筛选和交叉干扰下拉曼光谱非线性建模方法等主要内容的研究。.重要结果：1. 通过构建包含1600多种矿物的代表性拉曼光谱库，设计基于多尺度卷积神经网络结合通道注意力机制的复杂矿物组分快速识别模型，完成其在便携式拉曼光谱系统上的部署实现，实现了一套能够应用于矿物浮选过程中多组分快速精确鉴别的仪器设备，为浮选过程入矿性质的分析提供有力工具。2. 提出一种基于三次样条平滑结合离散状态转移算法的拉曼光谱强荧光背景估计及扣除方法，能够实现拉曼光谱中强荧光散射信号的消除，消除了拉曼光谱的浮选过程精矿品位检测的荧光散射信号影响。3. 提出了一种基于原位拉曼光谱技术结合CNN-LSTM模型的矿物浮选精矿品位预测方法，通过在线连续采集浮选泡沫层的拉曼光谱信号，采用CNN-LSTM网络充分利用光谱之间的时序信息实现精矿品位的预测，为矿物浮选过程中的精矿品位检测提供强有力工具。.科学意义：本项目针对矿物浮选过程中精矿品位难以在线检测的问题，提出了一种基于拉曼光谱的新的矿物浮选精矿品位在线检测方法，设计了相应的拉曼光谱建模方法，完成相应的算法模型训练和硬件系统设计，突破基于拉曼光谱的矿物浮选精矿品位在线检测关键技术，形成拉曼光谱应用于精矿品位在线检测的理论与方法，最终通过工业应用验证理论和方法的有效性。