Clustering process for analyzing measured data

摘要

Methods and a system for analysing measured data is provided. A measurement clustering method comprises obtaining multidimensional measurement data of an environment consisting of a plurality of variables, and obtaining at least one prototype 144 defining expected relationships between the variables for a particular property of the environment. The measurements in one or more clusters 146 are assigned to the prototype(s) and it is determined if the measurement clusters 146 fit the prototype(s) 144. The measurements are assigned to a new prototype if they do not fit and the acts of determining and reassigning are repeated until each of the prototype(s) fit a cluster. A physical property(s) of the environment is then predicted based upon the characteristics of the fitted prototypes. The method has particular application to is pressure gradient analysis in oilfield formation data.