神经网络和特征分析在矿产普查与勘探中的互补性探讨  

MUTUAL CONPENSATION OF NERVAL NET AND CHARACTERISTIC ANALYSIS IN MINERAL RESOURCES EXPLORATION

作　　者：董耀松[1] 杨言辰[2] 刘光胜[3] 夏立显[3] 

机构地区：[1]中国地质大学(武汉)地质过程与矿产资源国家重点实验室,湖北武汉430074 [2]吉林大学地球科学学院,吉林长春130026 [3]吉林大学综合信息矿产预测研究所,吉林长春130026

出　　处：《矿产与地质》2006年第1期1-6,共6页Mineral Resources and Geology

基　　金：中国地质调查局项目"华北地台成矿规律和找矿方向综合研究项目"(编号:200110200038)

摘　　要：近些年来,神经网络技术在矿产普查与勘探中被广泛应用,但是神经网络技术在成矿预测中存在明显的不足,它只是根据目标模式识别输入模式,而没有直接考虑变量和地质单元之间的关系,而且预测结果的可靠性很难评价。而特征分析方法是通过研究区内已知地质单元的研究,查明地质变量之间的内在联系并确定它们的找矿意义。因此,以华北地块北缘的铅锌矿成矿预测为例,把神经网络技术和特征分析结合起来进行成矿预测,可以弥补神经网络技术的不足,还可以提高预测成果的可靠性。这对找矿工作有很好的实际意义。The Nerval Net Technology is being widely applied to minerals exploration in recent years. But Nerval Net is obviously deficient in metallogenic prognosis, since they can＇t immediately calculate the relation between the variable and geological unit, but only identify the input mode according to the object pattern. It is very difficult to evaluate the result of metallogenic prognosis. But the characteristic analysis is to find out the inner relation among the geological variables and ascertain its prospecting significance by studying the known geological unit. Taking the Pb-Zn ore metallogenic prognosis of the northern margin of northern China platform as example, Nerval Net was integrated with characteristic analysis to remedy its deficiency. This method can also improve the reliability of the result of prognosis, It is of practical significance to prospecting work.

关 键 词：神经网络 特征分析 互补性探讨 铅锌矿 成矿预测 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]