基于赋存状态多维语义特征向量化的锂矿石相似性评价方法研究
杨佳伟 , 张康辉 , 高希宇 , 张晓亮 , 余瑾彦 , 朱阳戈
有色金属(选矿部分) ›› 2026, Vol. 0 ›› Issue (9) : 122 -131.
基于赋存状态多维语义特征向量化的锂矿石相似性评价方法研究
A Method for Ore Similarity Evaluation Based on Multidimensional Semantic Feature Vectorization of Occurrence States
随着矿产资源品位下降与矿石类型日益复杂,矿石赋存状态差异显著、矿物学特征维度多样,传统经验驱动的选矿工艺设计方法已难以满足资源高效利用的需求。针对复杂矿石赋存特征难以统一表达和定量比较的问题,本文提出基于赋存状态多维语义特征向量化的矿石相似性评价方法。首先,构建包含矿石类型、化学成分、矿物组成、嵌布特征及解离度等多维属性的结构化特征体系,并引入语义特征向量化方法,实现复杂矿物学信息的数字化与语义化表达。其次,基于词频-逆文本频率指数(Term frequency-inverse document frequency,TF-IDF)特征向量化与余弦相似度算法,建立矿石赋存特征的相似性评价模型,并通过人工设定的特征重要性权重调节机制提升关键赋存属性的区分能力。通过对待分析矿石样本与典型矿石样本库的语义相似度对比,实现不同赋存状态矿石在特征层面的相似性度量与筛选。研究结果表明,该方法可实现复杂矿石赋存特征的统一表征,能够从矿物学特征维度量化矿石间的相似程度;需注意矿石赋存特征相似并不直接等价于选矿工艺及分选指标一致,本研究可为后续开展选矿工艺知识迁移、工艺候选方案匹配与智能决策研究奠定数据与方法基础。
With the continuous decline in mineral resource grades and the increasing complexity of ore types, significant differences in ore occurrence states and the high dimensionality of mineralogical characteristics have made traditional experience-driven beneficiation process design methods insufficient to meet the requirements of efficient resource utilization. To address the difficulty of unified representation and quantitative comparison of complex ore occurrence features, this study proposes an ore similarity evaluation method based on multidimensional semantic feature vectorization of occurrence states. Firstly, a structured feature system is constructed by integrating multiple attributes, including ore type, chemical composition, mineral assemblage, textural characteristics, and degree of liberation. Semantic feature vectorization techniques are then introduced to achieve digitalized and semantic representations of complex mineralogical information. Subsequently, an ore similarity evaluation model is established using term frequency-inverse document frequency (TF-IDF) feature vectorization and cosine similarity, and an empirically-defined feature importance weighting adjustment mechanism is incorporated to enhance the discriminative capability of key occurrence attributes. By comparing the semantic similarity between target ore samples and a database of representative ore samples, similarity measurement and screening of ores with different occurrence states at the feature level are realized. The results demonstrate that the proposed method can realize unified representation of complex ore occurrence features and quantitatively evaluate ore similarity from mineralogical perspectives. It should be noted that similar mineralogical occurrence features do not guarantee consistent beneficiation performance. This work provides a data and methodological foundation for subsequent research on beneficiation process knowledge transfer, candidate-scheme matching, and intelligent decision-making.
| [1] |
俞柏. 战略性矿产资源供需形势分析与判断[J]. 中国金属通报, 2025(8): 11-13. |
| [2] |
|
| [3] |
陈从喜. 战略性矿产资源开发利用研究进展与展望[J]. 中国矿业, 2025, 34(2): 4-19. |
| [4] |
|
| [5] |
李鹏, 刘艳, 唐晨宇, |
| [6] |
|
| [7] |
王登红. 关键矿产的研究意义、矿种厘定、资源属性、找矿进展、存在问题及主攻方向[J]. 地质学报, 2019, 93(6): 1189-1209. |
| [8] |
|
| [9] |
|
| [10] |
苏妤芸, 苏秀珠, 刘全军, |
| [11] |
|
| [12] |
孙传尧, 周俊武. 流程工业选矿过程智能优化制造发展战略[J]. 有色金属(选矿部分), 2019(5): 1-5. |
| [13] |
|
| [14] |
吕燕, 周家云, 惠博, |
| [15] |
|
| [16] |
叶超, 罗政. 某低品位铌矿石工艺矿物学特征及可选性研究[J]. 中国矿山工程, 2021, 50(6): 87-91, 96. |
| [17] |
|
| [18] |
张愿宁, 郭园园, 逯文辉, |
| [19] |
|
| [20] |
朱汇派, 费光春, 谭华, |
| [21] |
|
| [22] |
汤化明, 郎增瑞, 王宏铭, |
| [23] |
|
| [24] |
蔡明明, 李光胜, 朱幸福, |
| [25] |
|
| [26] |
|
| [27] |
王智宇, 王达, 邱昆峰, |
| [28] |
|
| [29] |
田妞, 黄宋魏, 和丽芳, |
| [30] |
|
| [31] |
|
| [32] |
于潇然, 徐泉, 初延刚, |
| [33] |
|
| [34] |
于潇然. 选矿生产指标趋势监控与分析系统的设计与开发[D]. 沈阳: 东北大学, 2020. |
| [35] |
|
| [36] |
|
| [37] |
|
地球深部探测与矿产资源勘查国家科技重大专项(2024ZD1002000)
/
| 〈 |
|
〉 |