【作者】 刘永刚；

【导师】 杜德文； 石学法；

【作者基本信息】 中国海洋大学 ， 海洋地质， 2011， 博士

【摘要】 作为战略矿产资源,多金属结核、富钴结壳、热液硫化物等深海固体矿产越来越引起人们的关注,世界各国纷纷把战略目光转向了空间广阔、资源丰富的海洋,深海矿产资源的调查与评价工作越来越受到人们的重视。然而,由于深海和陆地矿产资源评价工作的的差异性,诸如海底调查程度低、控矿因素复杂且不确定性高、背景数据资料少且成矿模式不统一等,在陆地矿产评价中发挥重要作用的定量评价方法没有被系统地应用于深海矿产资源评价工作中。目前,深海矿产资源调查与评价工作是依据已有认识或专家判断,来选择调查区的,基本上未使用矿产资源定量评价技术来圈定重点勘探靶区。随着深海矿产资源调查技术的长足发展和资源调查工作的逐步深入,到了将陆地固体矿产资源定量评价理论和方法引入深海的时候了。本文拟将此作为研究工作的重点,尝试将陆地矿产资源定量评价方法引入深海,结合深海矿产资源评价特点,给出较系统定量评价方法。结合陆地矿产资源定量评价思路提出较系统的深海固体矿产资源定量评价方法,并将该方法应用于CC区锰结核资源。具体思路如下:首先,收集、整理和总结深海目标矿区(或典型矿区)的矿产数据资料及知识,收集、整理涵盖典型矿区的周边广大海底的背景数据资料及知识。对典型矿区前人研究成果的收集、整理和总结,可以得到成矿规律性的认识,为矿产的定量预测模型的建立,提供先验知识;对包含典型矿区的广大周边海底信息的收集整理,可以为典型矿区的预测模型推广提供证据图层。?其次,进行数据信息预处理。数据处理方法主要包括数据去异、数据网格化、数据变换和预测单元的选取等等。通过预处理,排除异常数据的影响,克服数据单位和量级不一致的所带来的弊病,同时将不规则数据变成均匀分布的数据矩阵,满足模型计算的需要。数据处理分析水平直接影响矿产资源定量预测建模的效果。?再次,对经过预处理后的数据进行信息特征提取。提取信息特征的方法有很多种,这里只介绍适合深海矿产资源相关数据的信息特征提取方法。这些方法包括空间场特征、统计特征、定性特征等的提取方法。信息特征提取技术能为矿产资源评价提供更多的证据图层,为矿产资源定量建模提供自变量信息。?最后,引入了适合实验区资源评价的两个新模型,即“证据权回归模型”和“Fuzzy?ARTMAP模型”。证据权回归模型可同时实现资源的定位和定量预测。改进的成分化Fuzzy?ARTMAP模型,能同时估算出多种金属的品位。前者是将两个已有算法进行组合使用而衍生的新方法,后者是第一次被应用于矿产资源定量评价工作中。这两个模型在实验区的实际应用中取得良好效果,被证明适合深海矿产资源定量评价工作。?更多还原

【Abstract】 As strategic mineral resources, deep‐sea solid mineral resources, such as polymetallic nodules, cobalt‐rich crusts, and hydrothermal sulfide, has drawn more attention and many countries have shifted their strategic vision to the vast and resource‐rich ocean, so deep‐sea mineral resources surveying and assessment become more and more important. However, due to the differences between deep‐sea and land mineral resources, such as the low level of seabed surveying, complex ore‐controlling factors and high uncertainty, lack of background data various metallogenic model of data and so on, quantitative estimation methods, playing an important role in land mineral assessment, has not been systematically applied in the estimation of deep‐sea mineral resources.At present, the deep‐sea mineral resources surveying and estimation is based on the existing knowledge or expert judgments to select the survey area, basically not using quantitative assessment technology to delineate target areas. With the development of deep‐sea mineral resources surveying techniques, it is time to introduce the theory and methods of land mineral resources quantitative evaluation to deep sea. This research work intends to focus on this point trying to introduce the theory and methods of land mineral resources quantitative evaluation to deep sea, and gives a more systemic quantitative assessment method, combining with the characteristic of deep‐sea mineral resources assessment. The systemic deep‐sea mineral resources quantitative assessment methods are proposed based on the thoughts of land solid mineral resources quantitative assessment, and are applied to manganese nodule resources in CC zone. Specific ideas are as follows:First, collect, classify and summarize the mineral data and knowledge in deep‐sea target mining area (or typical mining area), then collected, classify the background data and knowledge on typical mining area and its surrounding areas. The knowledge of metallogenic regularity can be obtained by collecting, classifying and summarizing the results of previous studies on typical mining area, to provide a priori knowledge for establishing the estimated model. The collection and classification of the information on the typical mining area and its surrounding areas can provide evidence layers for generalization of estimated model on typical mining areaSecond, preprocess data. Data processing methods include removing abnormal data, data gridding, data transformation and estimating unit selection and so on. Through the preprocessing, the impact of abnormal data is excluded, the inconsistencies of data units and magnitude is overcome, while irregular data is transformed into a uniform the data matrix, meeting the needs of model calculation. Direct impact on the level of data processing and analysis impacts directly the effect of modeling.Third, feature extraction after data preprocessing. There are many ways for feature extraction, and here only the information extraction methods for deep‐sea mineral resources data are introduced,. These include methods for extracting spatial field features, statistical features and the qualitative fractures. Information extraction technology can provide more evidence layers and independent variable information.Finally, introduce two new assessment models for experiment area. They are "weight of evidence regression model" and "Fuzzy ARTMAP model." Weight of evidence regression model realize simultaneously positioning and quantitative estimation. Modified Fuzzy ARTMAP element model, which can estimate the metallic grade of many kinds of metal. The former is a combination of two existing algorithms, and the latter is used in the work for the first time. The two models get good results in the practical application, which proves to be suitable for quantitative assessment of deep‐sea mineral resources.更多还原