模糊聚类分析在地层对比中的应用

模糊聚类分析已经广泛地用于油气检测和构造评价,它也可以用于地层对比。以往在煤田勘探中,多采用相关法进行地层对比,即以一已知钻井的测井曲线作为对比标准,与待对比的测井曲线进行相关运算,以相关系数的大小作为确定地层对比关系的依据。实际上相关分析的结果是模糊的。从模糊数学的观点看,当作对比标准的曲线可以视为“模式”,地层相关对比过程是“模式识别”过程。由此可知,地层对比可以看成是-个模糊聚类分析问题。典型的系统聚类分析法以及动态聚类分析法对地层对比是不适用的。这里介绍的方法是这样的:(1)首先计算k地层的测井曲线与待对比的测井曲线的相关系数;(2)综合各类测井曲线的相关系数作为子样的隶属函数;(3)初步选定隶属函数的阈值,若隶属函数值大于或等于选定的阈值,则初步认定在待迷对比的钻井井段找到了相应k地层的曲线,初选的阈值不合适也无妨,还可反复修改;(4)将上述认定属于k地层的曲线再与待对比的测井曲线段进行上述运算,又获得-批新的归属k地层的曲线段,以此新曲线段为准,再重复上述步骤,直到再无新的属于k地层的曲线段为止。对k=1,2,…,K 等情况均进行相同的处理,即对初步选定的阈值完成了具有k个子集的聚类分析。

Application of fuzzy cluster to stratigraphic correlation

Fuzzy cluster analysis has been used in hydrocarbon indication and trap evaluation.It also can be applied to stratigraphic correlation. In coal exploration, strata arc generally correlated by correlation method.According to correlation method, a known logging curve is used as a standard and correlated with a new logging curve so as to get their correiaiion coeffieient, which is then taken as the basis of stratigraphic correlation. Actually, the result of such correlation analysis is fuzzy.If viewed from fuzzy mathematics,the known logging curve is used as a "pattern", and the correlation of strata is called "pattern identificanon". It follows that stratigraphic correlation can be considered as a fuzzy cluster analysis questio. It must be pointed out that the typical systematic cluster analysis and dynamic cluster analysis are not usable to stratigraphic correlation.A useful method is recommended here and its procedure is as follows:(1)computing the correlation coefficients between the known logging curves of the kth stratum and the curves to be correlated.(2)obtaining the weighted average of correlation coefficients,which is then used as the membership function of subset.(3)choosing the preliminary threshold value of the membership function. When a functional value is not less than the threshold value, we think it possible that the curve segment corresponding to the kth stratum is found. It doesn't matter whether the preliminary threshold value is suitable or not, because it can be modified again and again.(4)the found curve segment corresponding to kth stratum is then used to correlated with another new set of logging curves, we then obtain another kth stratum curve, which can be further used to repeat the process. The repeated process goes on until the end.After all the processings of k strata (k=1, 2, 3,…,K) are done,we can say that the cluster analysis of k subsets for the preliminary threshold is finished.