Multivariable regression of thermal conductivity in rocks
Jong Jeong Yeon, Sup Yun Tae, Yeom Kim Kwang · IOS Press eBooks · 2015
This study introduces multivariable regression of thermal conductivity based on statistical correlation with geophysical properties of rocks which are compressional wave velocity, shear wave velocity, density, porosity, and mineral compositions (quartz contents, plagioclase contents, feldspar contents, and hornblende contents). Collected 42 rocks in South Korea, which are composed of igneous, metamorphic, and sedimentary rocks, are subjected to laboratory experimental measurement of compressional and shear wave velocities, density, porosity and mineralogy. Considered geophysical properties are evaluated based on the assessments of Multivariable Linear Regression (MLR) analysis and coefficient of determination (R2) values. Among variables that are inter-correlated and dependent under consideration, the compressional wave velocity is statistically meaning and governing properties to estimate thermal conductivity. The Artificial Neural Network is additionally implemented to achieve the accurate prediction with varying the number of hidden layers and neurons.