Computer Application of Principal Component Analysis to Boundary Identification and Well-to-Well Correlation

Saleh, Mohamed Rahuma (1991) Computer Application of Principal Component Analysis to Boundary Identification and Well-to-Well Correlation. MSc(R) thesis, University of Glasgow.

Full text available as:
[thumbnail of 11011418.pdf] PDF
Download (9MB)


Many studies have employed automated procedures in well-log interpretation to aid the identification of formation boundaries, and perform cross-correlation between formations in different boreholes. These investigations have demonstrated the use of only one well-log variable, usually resistivity or Gamma Ray logs in the process. In conventional well-log interpretation, different well-log variables or a group of variables are used for different tasks. This project makes use of principal components and spectral analysis as the basis for well-log interpretation, including automatic formation boundary identification and cross-correlation using the first principal component of well-log variables. By transforming a set of well-log data using principal components analysis a single new variable is extracted from the first principal component scores which accounts for a significant amount of the variation within the original data. A further improvement in the results is obtained by passing the data through a moving filter to reduce noise. Boundary identification is performed by generalized distance (D2 ) method. Cross-correlation between the filtered principal components of two boreholes is then made by matching each formation of one borehole with a part of another. Both the stretch factor which accounts for thickening or thinning between sequences and the relative vertical displacement of the formation are calculated. This requires the calculation of the power spectra, derived form the fast Fourier transform of the principal component data, with high pass filtering using the derivative filter to obtain the appropriate resolution. This new technique was applied on model and real well-log data from five boreholes in the Attahaddy field, Libya. Although the Attahaddy field is structurally complex, the method was found to be reliable at predicting both the geological boundaries of the different formations, and the correlation of formations between boreholes. The distinctive value of this new approach is in its application of the first principal component of the original well-log variables. Such application has many advantages over the previous studies. PCAXCOR is a new computer program written in F77 to perform all the necessary computation for boundary identification and well-to-well correlation based on principal component analysis. Graphical output of the results uses a number of new functions in the S language.

Item Type: Thesis (MSc(R))
Qualification Level: Masters
Keywords: Geophysical engineering, Petroleum engineering
Date of Award: 1991
Depositing User: Enlighten Team
Unique ID: glathesis:1991-78351
Copyright: Copyright of this thesis is held by the author.
Date Deposited: 28 Feb 2020 12:09
Last Modified: 28 Feb 2020 12:09

Actions (login required)

View Item View Item


Downloads per month over past year