مهندسی عمران شریف (Aug 2017)
COMPARISON OF DIFFERENT FAILURE CRITERIA IN ROCK MASSES USING OPTIMIZATION (TS) AND CURVE FITTING
Abstract
Failure criterion has long been known and used as a useful criterion for evaluating the strength properties of rocks. Failure criteria development process and increase of their accuracy in predicting the lateral strength of rocks under pressure indicate the high performance and precision of modern standards in evaluating features of rock strength. The development of failure criteria begins with the initial theoretical criteria and leads to empirical criteria based on curve fitting on tri-axial failure data. Accurate theoretical criteria are not reliable in predicting the rock strength regarding most natural stones, and it is proved to be due to their nature.Although theoretical criteria are critical and necessary for better comprehension of rock behaviour, their application are very limited in practice. Designs' experimental criteria are more frequently used. In recent five decades, in order to simulate tri-axial behaviour of rock specimen, several experimental criteria have been presented which only a few of them have gained popularity. Function forms in these criteria have been different: separate constants are introduced for estimation of the strength of the rock type in every one of them. In this study, the TABU Search (TS) with its considerable features that can help us as a powerful tool in the optimization of difficult problems is used to optimize the mentioned criteria. To use TABU search techniques in this study, the algorithm is written in MATLAB.In this paper, Hoek---Brown, Fairhurst, Franklin and Bieniawski criteria have been selected as empirical criteria. Verifying the predictions of the four criteria concerning the experimental data of different rock samples will be analyzed; moreover, their advantages will be compared. In this study, constants for each criterion are obtained using two methods; the curve fitting and the optimization with TABU's algorithm. The results achieved in this study indicate that the use of optimization algorithms, compared with the curve fitting (regression), can have considerably better envelope failure prediction for different rocks.
Keywords