Assessment of Deformation Properties of Coal Measure Sandstones Through Regression Analyses and Artificial Neural Networks
Özet
The deformation properties of rocks play a crucial role in handling most geomechanical problems.
However, the determination of these properties in laboratory is costly and necessitates special equipment.
Therefore, many attempts were made to estimate these properties using different techniques. In this study,
various statistical and soft computing methods were employed to predict the tangential Young Modulus
(Eti, GPa) and tangential Poisson’s Ratio (vti) of coal measure sandstones located in Zonguldak Hardcoal
Basin (ZHB), NW Turkey. Predictive models were established based on various regression and artificial
neural network (ANN) analyses, including physicomechanical, mineralogical, and textural properties of
rocks. The analysis results showed that the mineralogical features such as the contents of quartz (Q, %) and
lithic fragment (LF, %) and the textural features (i.e., average grain size, d50, and sorting coefficient, Sc)
have remarkable impacts on deformation properties of the investigated sandstones. By comparison with
these features, the mineralogical effects seem to be more effective in predicting the Eti and vti. The performance of the established models was assessed using several statistical indicators. The predicted results
from the proposed models were compared to one another. It was concluded that the empirical models
based on the ANN were found to be the most convenient tools for evaluating the deformational properties
of the investigated sandstones.
