Study On Strata Multivariate Statistical Analysis Methods Based On Matlab

Regarded Matlab as a platform and Guided by sequence theories, this article adopts a series of methods such as wavelet transform, principal component analysis, clustering analysis and so on to deal with well-logging data and fully dig geological information which implies in well-logging information to study wavelet analysis pattern of sequence unit and multivariate statistical pattern of rock stratigraphic unit. In this article, Matlab is used to compile programme. By doing this some functions such as interpolation, principal component analysis, wavelet transform and clustering analysis are successfully realized. We use interpolation module acquired by compiling pogramme to interpolate typical well-logging curves. By comparing the interpolating results that adopt different interpolation parameters we consider linear as interpolation parameter. By making use of wavelet transform function in the compiled programme to analyze typical well-logging curves it shows that it is favorable to divide sequence units when dimension is 81 and the maximum value is one-eighth of well-logging data number; It is favorable to divide subquence units when dimension is 21 and the maximum is one-tenth of well-logging data number. After using the PCA module in the programme to analyze well-logging data, the validity of regarding the first principal component as the object of wavelet transform can be proved. Combined with the analysis results of typical well-logging curves, seven wells in Jiyang-Linqing area are used to divide sequences and subsequences. In researching multivariate statistical pattern of rock stratigraphic units it mainly adopts the clustering analysis module, which includes hierachical clustering analysis module and K-means clustering analysis module. By dealing lithology of Shanxi and Taiyuan Formation of seven wells in Jiyang-Linqing area with clustering analysis we can get some multivariate statistical patterns of rock stratigraphic units.That is to say, for Shanxi Formation when cluster number is 6 the values of hierarchical clustering intercluster distance are between 5 and 7 while those of Taiyuan are between 3 and 7. After doing K-means clustering analysis it shows that for Shanxi Formation the silhouette average values are not the minimum among all clusters' silhouette average values when k is 4. And those average values are not the maximum while those of Taiyuan Formation are not the minimum among all clusters' silhouette averager values when k is 3. This kind of pattern is verified in analyzing DG2 and get better effect.