Application of Soft Computing Techniques for Prediction of Slope Failure in Opencast Mines

One of the most arduous jobs in the industry is mining which involves risk at each working stage. Stability is the main focus and of utmost importance. FOS when calculated by traditional deterministic approach cannot represent the exact state at which the slope exists, though it gives a rough idea of the conditions and overall safety factor. Various approaches like numerical modelling, soft computing techniques allow us with the ease to find out the stability conditions of an unstable slope and the probability of its failure in near-by time. In this project, the stability conditions of some of the benches of Bhubaneswari Opencast Project, located in Talcher, have been evaluated using the soft-computing techniques like Artificial Neural Network implemented using MATLAB and then the results are being compared with the Numerical Model results from the software FLAC which deploys Finite Difference Method. A particular slope (CMTL-179, Seam-3) has been studied and the respective factor of safety for each slope has been predicted using both the Artificial Neural Network and FLAC. Initially the data related to bench height, slope angle, lithology, cohesion, internal angle of friction, etc. are determined for the respective rock of the slope of which the FOS is to be calculated. . A total of 14 training functions were used to train the model. The best training was found in Scaled Conjugate Gradient Backpropagation which corresponds to a regression coefficient of 91.36% during training and 88.24% overall. The best Validation Performance was also found at 60 epochs with Mean Squared Error of 0.069776. According to the trained neural network, it was found that the slope was 44.5% stable with a FOS 1.0226. Using the software FLAC, it was found that the slope was stable with FOS=1.17. The generic model will thus allow us to get a range of probability for the slope to fail so that necessary arrangements can be made to prevent the slope failure