No abstract

No abstract

In-situ stress is one of the most important input data to study stability analysis of underground and surface geomechanical projects. The measured vertical stress has a linear relation with depth. The average value of unit weight (g) was obtained 0.026 MN/m3 (2.56 ton/m3) using 1041 test results of different rocks with 0.001 difference with 0.027 MN/m3 which is a reliable coefficient for estimating vertical stress. The ratio of horizontal (sh) to vertical (sv) stress (K =sh/sv) is estimated by theoretical and empirical methods. The results showed that the estimating ratio of horizontal to vertical stress (K) by a theoretical method such as Terzaghi and Richard is much smaller than 1, and the estimation of the K value utilizing empirical methods such as Hoek and Brady is much greater than 1, with its value even approaching 4 in the near ground surface. To overcome the lack of an applicable comprehensive relation for the estimation of the K ratio and improve the shortcomings of previous methods, a new empirical relation was developed to estimate the K ratio utilizing a significant number of in-situ test results. Stability analysis of Masjed Soleyman powerhouse caverns was carried out by numerical modelling for five values of the K ratio obtained by previous stress estimation methods and this study. The in-situ stress estimation method (K ratio changes) showed a significant effect on stresses, displacements, strains, depth of the plastic zone and significantly affect the stability analysis and support system design of the powerhouse and transformer caverns.

No abstract

A gene expression programming (GEP) model has been used for predicting the risk of building damage due to ground movement induced by tunnelling. A critical displacement parameter, namely deflection ratio, and building major affecting factors have been identified from the previous empirical method. A general GEP model is trained and evaluated using the actual collected data from Karaj Railway in Iran, and published case histories from other projects around the world. The general model results are compared with the actual field measurements for the training of the model purpose. The trained network is used to estimate the risk of damage for a few case histories. It is shown that not only a GEP model learns a relationship between the risk of building damage with appropriate ground movement and building parameters, but it can also develop the learning to predict damage for very different geological and tunnel and building geometric conditions.