Document Type : Research Note
Authors
1
School of mining, College of engineering University of Tehran
2
Dept. of Mining Engineering University of Kashan
Abstract
A grouting operation is one way to decrease water penetration, and increase the resistance and strength of jointed rocks in locations of dams. One encountered problem in this operation is evaluation of consumed cement and estimation of the penetration and radiuses of cement grout. In this study, analytical models with numerical modeling and a neural network to predict the penetration radius and the volume of grout are introduced.To predict the grout volume required for grouting and penetration radius, simplifying the geometrical state of the rock and penetration conditions of the grout is necessary. The cement grout used in the grouting operation is known as a Bingham flow. Thus, grout rheological properties, including viscosity, $\mu$ properties, including viscosity, , and yield stress, $\tau_0$, will control its behavioral properties. Viscosity, flow speed of the grout and yield stress will control the most penetrating length of the grout in boring in defined grouting pressures and the constant aperture of joints.Analytical models for prediction of cement grout have been used in this study. In numerical modeling by UDEC software, the conditions of the rock, joint apertures, and grout bore dimensions and grout pressure values are entered. Because of the low regression coefficient between real and calculated grout volume, a neural network method is used to present better predictions. The purpose of this study is the introduction of calculation methods for evolution of grout volume by these methods. Finally, modeling was undertaken using data from the Seimareh dam, and the obtained grout volume was compared to the real
take recorded in this site.Using analytical models, the required grout volume can be estimated. The fourth model (Lombardi), in comparison to other models, presents a better estimation of grout volume. Numerical modeling, using UDEC software, in comparison to analytical models, has better estimation of grout volume, and the calculated take regression with real take is higher. But, the amount of regression coefficient is not enough $(R^2= 0.628)$. Stimulation results and calculated values in the neural network method by MATLAB software present the best estimation of grout volume and acceptable regression coefficient $(R^2=
0.92)$. Thus, using a neural network system to predict cement grout take in the grouting process is presented.
Keywords
)
