Prediction of cementitious grout take for a mine shaft permeation by adaptive neuro-fuzzy inference system and multiple regression

Abstract

Cement grouting is a common technique implemented for permeation and ground improvement in civil and mining engineering projects. Basically, it is the injection of cement and water mixture into a fractured rock mass. Due to the presence of water bearing and permeable rock mass, permeation grouting was applied prior to the shaft sinking operation in an underground mine, located in Soma coal basin, Turkey. The Drill-Grout-Drill (DGD) method was used in permeation grouting for a flood prone mine shaft project with a circular pattern covering the proposed shaft opening. Data collection was mainly based on recording borehole data, however, during shaft sinking, field observations were continued to check and validate data, especially the rock mass properties. Widely used classification systems, such as RQD and RMR discontinuity condition rating were selected to define rock mass parameters. The rock mass parameters and the grout take data were pre-processed and cleaned to be used as input for multiple regression modelling and Adaptive Neuro Fuzzy Inference System (ANFIS). Linear, nonlinear, and Box-Cox multiple regression models provided accurate results. ANFIS with subtractive clustering and with manual dictation resulted in improved predictions compared to the regression analysis. Since grouting has great complexity and dependence on numerous variables, particular limitations and omissions had to be defined within the scope of the research. All influential factors could not be interpreted. The methodology and variable conditions are the main novelties of this study and enhance the implementation of the method specifically in the mine project where the study was carried out.