Empirical relations between compressive strength and microfabric properties of amphibolites using multivariate regression, fuzzy inference and neural networks: A comparative study

• Quantitative image analysis of rocks is an acceptable approach in the initial design stage.
• Empirical relationships to estimate UCS of rocks from microfabrics properties is more practical.
• Grain size, shape factor and quartz % are tailored to estimate UCS of banded amphibolite rocks.
• Neural networks method is best model for the reliable assessment of compressive strength
• Predictive capability of regression model is quite accurate than fuzzy inference system.

In this paper, microfabric properties including grain size, shape factor and quartz content are tailored to the specific evaluation of UCS of banded amphibolite rocks. However, the predicting capabilities of Artificial Neural Networks (ANNs) and Fuzzy Inference System (FIS) as well as the Multivariate Regression (MR) techniques have been evaluated and compared using the same input variables. To assess the model performances, some performance indices such as correlation coefficient (R), variance account for (VAF) and root mean square error (RMSE) were calculated and compared for the three models. The study revealed that even though the developed three models are reliable to predict the UCS, the presented ANN method displays an obvious potential for the reliable assessment of UCS according to model performance criterion. However, the outcomes of this study are quite satisfactory, which may serve microfabric characterization to be easily extended to the modeling of strength and deformation behavior of rocks in the absence of adequate budget and facility of testing UCS.