In the paper published by us  we set forth the methods and results of multivariate formation classification using the data base of geophysical parameters.
Four types of formations have been identified with a 87–94% accuracy presenting non-reservoirs, water saturated and oil saturated reservoirs and one class which we called intermediary or mixed which judging by the results of standard conclusions on well-logging data, includes in various proportions the formations of all the above-mentioned classes. The present paper illustrates the results of classification with identification of other 6 or 8 groups of formations with the purpose of their detailization according to lithology and fluid saturation.
The used data base included 6 geophysical parameters characterizing 14,546 formation intersections in 1,707 wells. The classification method lies in building of a ‘Tree of Classes’ according to the Ward method and appropriate classifying equations of discriminant analysis, and in determining the position of classes on the ordination plot in compliance with canonical equations.
The classification results in the delivery of:
Borders of classes identified implicitly by means of discriminant analysis and described by classifying functions;
Tables with the results of classification quality on the training subsets, including the percentage of correct attribution of the interval to the class;
Statistical description of classes.
Even though interpretation results were not used when making up classifying equations, the comparative analysis of classification results with standard interpretation data points to possibility and practicability of including the generated classifying and canonical equations into the active software for interpretation of well-logging data.
This will increase accuracy and unambiguity of issued conclusions on the lithology of formations and their fluid saturation, which is very important for building geological and hydro-dynamic models of oil fields and estimation of reserves and designing of development.
The survey was targeted at the well-logging results acquired from wells in the Devonian deposits in two areas of the Romashkinskoye field which has been under development since the fifties of the last century. Drilling out of Devonian terrigenous pay zone is done continuously as the geological structure of the object is being defined more precisely in order to perfect the system of the oil field development .
Building of the model involved the use of data base with geophysical parameters PZ, PS, ANG, AG, IK and BK, which were previously transformed and standardized (Table 1).